PLplot Release 5.11.1
This is a release of the PLplot plotting package. It represents the
ongoing best efforts of the PLplot community to improve this package,
and it is the only version of PLplot that we attempt to support.
Releases in the 5.x.y series should be available several times per
year.
If you encounter a problem with this release that is not already
documented on our bug tracker, then please send bug reports to PLplot
developers via our mailing lists (preferred for initial discussion of
issues) at <http://sourceforge.net/p/plplot/mailman/>. If it turns out
no quick resolution is possible via mailing-list discussion, then the
issue should be placed on our bug tracker at
<http://sourceforge.net/p/plplot/bugs/>.
Please see the license under which this software is distributed
(LGPL), and the disclaimer of all warranties, given in the COPYING.LIB
file.
________________________________________________________________
CONTENTS
1. OFFICIAL NOTICES FOR USERS
1.1 Minimum CMake version bump
2. Improvements relative to the previous release
2.1 Bug fixing!
2.2 Improve traditional build linking for nondynamic device drivers case
2.3 Comply with a NEW value for the CMake CMP0022 and CMP0023 policies
2.4 Comply with a NEW value for the CMake CMP0026 policy
2.5 Plot labels as date / time are now available for 3D plots
2.6 Comprehensive testing script improvements
2.7 Build system and file cleanup concerning retired devices
2.8 Many improvements and bug fixes for the new wxwidgets device
2.9 Substantially improved results for comprehensive testing of PLplot
on the Cygwin platform
2.10 Initial success of comprehensive testing of PLplot on the MSYS2
platform
3. PLplot comprehensive testing reports
________________________________________________________________
1. OFFICIAL NOTICES FOR USERS
1.1 Minimum CMake version bump
The minimum version of CMake has been bumped to 3.2.3 on all platforms
other than Cygwin and Linux where it has been bumped to 3.0.2 instead.
________________________________________________________________
2. Improvements relative to the previous release
2.1 Bug fixing!
Please consult the ChangeLog.release file for details concerning the
numerous bug fixes made for this release.
2.2 Improve traditional build linking for nondynamic device drivers case
For -DENABLE_DYNDRIVERS=OFF, the C and C++ code for the device drivers
becomes part of libplplot (as opposed to the -DENABLE_DYNDRIVERS=ON
case where the device drivers are independently built as dll's which
are dynamically loaded by libplplot). Thus, -DENABLE_DYNDRIVERS=OFF
makes libplplot a mixed C and C++ library that requires linking using
all libraries that are required by the C++ compiler. Which can get
complicated for the C, Fortran, D, Ada, Java, and OCaml examples.
CMake takes care of such linking issues automatically, and for the
traditional build of the installed examples we have replaced the
previous fragile and g++-specific scheme for determining the C++
compiler libraries with CMake's knowledge of those same libraries
for each C++ compiler supported by CMake.
2.3 Comply with a NEW value for the CMake CMP0022 and CMP0023 policies
All use of the deprecated LINK_INTERFACE_LIBRARIES has been replaced
by using either the PRIVATE (NON_TRANSITIVE=ON,
INTERFACE_LINK_LIBRARIES set to empty) or PUBLIC (NON_TRANSITIVE=OFF,
INTERFACE_LINK_LIBRARIES set to listed libraries)
target_link_libraries keywords. The PRIVATE or PUBLIC keywords were
also appropriately propagated to our configured pkg-config files and
thus to our traditional build system for the installed examples.
2.4 Comply with a NEW value for the CMake CMP0026 policy
This NEW CMake policy means that location properties must no longer be
used to determine the filenames of executables, dll's, and libraries
for our CMake-based build system. Instead, the same information must
be determined from CMake generator expressions. Complying with this
policy makes our build system much less sensitive to potential
location bugs at generate time but has required intrusive but
well-tested changes to our CMake-based build system.
2.5 Plot labels as date / time are now available for 3D plots
The plbox3 function now supports the d suboption for xopt and yopt and
the e suboption (d is already used for something different for zopt)
for zopt to generate X, Y, or Z axis labels using date/time
formatting. As in the 2D case (see standard example 29) call
the pltimefmt routine beforehand to control the date/time format that
is used.
2.6 Comprehensive testing script improvements
The principal improvement is to collect the most important information
concerning a given comprehensive test into a report tarball that helps
others to diagnose any problems turned up by the test or in the case of
complete test success provide enough information for a summary of the
test (see, for example, the tables below summarizing comprehensive
test success on various platforms).
2.7 Build system and file cleanup concerning retired devices
We have long since retired the gnome, gcw, gnome2, dg300, hpgl,
impress, linuxvga, ljii, ljiip, pbm, and tek device drivers. This
(intrusive!) change removed all references to these from files (or
removes entire files if the file was only relevant to one of the
retired device drivers). The only exception to this cleanup is our
DocBook documentation of devices which needs a complete rewrite
instead.
2.8 Many improvements and bug fixes for the new wxwidgets device
The new wxwidgets device is still being actively developed. See the
ChangeLog.release file for the details of all such changes for this
release cycle. The current status is this new version of the
wxwidgets device has no serious build or run-time issues. However,
there are still known rendering issues with this device which can be
found at <http://sourceforge.net/p/plplot/bugs/search/?q=wxwidget> and
which we are in the process of fixing. If any of these issues are an
important concern to our wxwidgets users for this release we suggest
they try the -DOLD_WXWIDGETS=ON option to access the old version of
the wxwidgets device. We do sufficient maintenance of that old
wxwidgets device so that it passes comprehensive testing (see the
report below). However, there are no plans to do further changes to
the old wxwidgets device driver other than minimal maintenance, and at
some point that old version of the wxwidgets device will be removed.
2.9 Substantially improved results for comprehensive testing of PLplot
on the Cygwin platform
See the detailed reports for 64-bit Cygwin below. A substantial
amount of bug fixing for our build and test system during this release
cycle contributed to this 64-bit Cygwin success. Because of this
success and because this platform provides almost every PLplot
soft prerequisite we highly recommend 64-bit Cygwin as a PLplot
platform with the caveat that the results depend on the Cygwin dll,
i.e., they are not pure Windows results.
We similarly expect that 32-bit Cygwin will be a good platform for
PLplot, but we have not comprehensively tested this platform yet.
2.10 Initial success of comprehensive testing of PLplot on the MSYS2
platform
Greg Jung reports that comprehensive testing of PLplot on
mingw-w64-x86_64/MSYS2 succeeds (see the detailed report below). This
derivative of the modern Cygwin platform allows use of posix MSYS2
tools to build pure (i.e., no dependence on the MSYS2 dll) Windows
applications and libraries using the mingw-w64-x86_64 toolchain and
the associated mingw64 software repository. Although this repository
does not contain all the free software that is available for Cygwin,
it nevertheless does cover most of the PLplot prerequisites. So we
can now highly recommend this platform for pure Windows builds of
PLplot using the mingw-w64-x86_64 toolchain. Thanks, Greg!
We similarly expect that the mingw-w64-i686/MSYS2 platform
will also work well for pure Windows builds of PLplot using the
mingw-w64-i686 toolchain although we haven't comprehensively tested
this platform yet.
________________________________________________________________
3. PLplot comprehensive testing reports
These reports are in reverse chronological order and formatted in the
SourceForge markdown syntax for a table.
|||
---|---|---|---
Tester|Alan W. Irwin
Notes|(a), (b), (c), (d), (e), (k), (A)
Date|2015-08-07
PLplot commit|6747256
CMake version|3.0.2
Generator|\"Unix Makefiles\"
Platform|Debian wheezy = oldstable with system libraries
Pango/Cairo version|1.30.0/1.12.2
Qt version|4.8.2
WxWidgets version|2.8.12.1
Shared libraries?|Dynamic drivers?
Yes|Yes
Yes|No
No|No
CMake-based build tree?|test_noninteractive?|test_interactive?|ctest?
Yes|No|Yes|No
CMake-based installed examples?|test_noninteractive?|test_interactive
Yes|No|Yes
Traditional Installed examples?|test_noninteractive?|test_interactive
Yes|No|Yes
|||
---|---|---|---
Tester|Alan W. Irwin
Notes|(a), (b), (c), (d), (e), (A)
Date|2015-07-30
PLplot commit|d64d9c6
CMake version|3.0.2
Generator|\"Unix Makefiles\"
Platform|Debian wheezy = oldstable with system libraries
Pango/Cairo version|1.30.0/1.12.2
Qt version|4.8.2
WxWidgets version|2.8.12.1
Shared libraries?|Dynamic drivers?
Yes|Yes
Yes|No
No|No
CMake-based build tree?|test_noninteractive?|test_interactive?|ctest?
Yes|Yes|Yes|Yes
CMake-based installed examples?|test_noninteractive?|test_interactive
Yes|Yes|Yes
Traditional Installed examples?|test_noninteractive?|test_interactive
Yes|Yes|Yes
|||
---|---|---|---
Tester|Greg Jung
Notes|(a), (e), (i), (j), (A), (B'), (E''''), (F'''), (G'), (H'''), (N)
Date|2015-07-30
PLplot commit|be85695
CMake version|3.2.3
Generator|\"MSYS Makefiles\"
Platform|mingw-w64-x86_64/MSYS2
Pango/Cairo version|1.37.1/1.14.2
Qt version|4.8.7
WxWidgets version|3.0.2-5
Shared libraries?|Dynamic drivers?
Yes|Yes
Yes|No
No|No
CMake-based build tree?|test_noninteractive?|test_interactive?|ctest?
Yes|Yes|No|Yes
CMake-based installed examples?|test_noninteractive?|test_interactive
Yes|Yes|No
Traditional Installed examples?|test_noninteractive?|test_interactive
Yes|Yes|No
|||
---|---|---|---
Tester|Greg Jung
Notes|(a), (e), (i), (A), (B'), (E''''), (F'''), (G'), (H''), (N)
Date|2015-07-30
PLplot commit|be85695
CMake version|3.2.3
Generator|\"Unix Makefiles\"
Platform|mingw-w64-x86_64/MSYS2
Pango/Cairo version|1.37.1/1.14.2
Qt version|4.8.7
WxWidgets version|Not applicable
Shared libraries?|Dynamic drivers?
Yes|Yes
Yes|No
No|No
CMake-based build tree?|test_noninteractive?|test_interactive?|ctest?
Yes|Yes|No|Yes
CMake-based installed examples?|test_noninteractive?|test_interactive
Yes|Yes|No
Traditional Installed examples?|test_noninteractive?|test_interactive
Yes|Yes|No
|||
---|---|---|---
Tester|Greg Jung
Notes|(a), (b), (e), (A), (B'), (D'), (E'''), (G'), (H'), (M)
Date|2015-07-11
PLplot commit|dd9f79e
CMake version|3.1.2
Generator|\"Unix Makefiles\"
Platform|Cygwin (64-bit)
Pango/Cairo version|1.36.8/1.12.18
Qt version|4.8.7
WxWidgets version|3.0.2
Shared libraries?|Dynamic drivers?
Yes|Yes
Yes|No
No|No
CMake-based build tree?|test_noninteractive?|test_interactive?|ctest?
Yes|Yes|No|Yes
CMake-based installed examples?|test_noninteractive?|test_interactive
Yes|Yes|No
Traditional Installed examples?|test_noninteractive?|test_interactive
Yes|Yes|No
|||
---|---|---|---
Tester|Arjen Markus
Notes|(a), (h), (A), (B'), (C'), (D), (D'), (E), (F'), (G), (H), (I), (J)
Date|2015-07-10
PLplot commit|32be1b9
CMake version|3.2.2
Generator|\"MSYS Makefiles\"
Platform|MinGW/MSYS
Pango/Cairo version|Not Applicable
Qt version|Not Applicable
WxWidgets version|Not Applicable
Shared libraries?|Dynamic drivers?
Yes|Yes
Yes|No
No|No
CMake-based build tree?|test_noninteractive?|test_interactive?|ctest?
Yes|Yes|No|Yes
CMake-based installed examples?|test_noninteractive?|test_interactive
Yes|Yes|No
Traditional Installed examples?|test_noninteractive?|test_interactive
No||
|||
---|---|---|---
Tester|Arjen Markus
Notes|(a), (b), (e), (A), (B'), (D'), (E''), (G'), (H')
Date|2015-07-07
PLplot commit|ae0e9da
CMake version|3.1.2
Generator|\"Unix Makefiles\"
Platform|Cygwin (64-bit)
Pango/Cairo version|1.36.8/1.12.18
Qt version|4.8.7
WxWidgets version|2.8.12.1
Shared libraries?|Dynamic drivers?
Yes|Yes
Yes|No
No|No
CMake-based build tree?|test_noninteractive?|test_interactive?|ctest?
Yes|Yes|No|Yes
CMake-based installed examples?|test_noninteractive?|test_interactive
Yes|Yes|No
Traditional Installed examples?|test_noninteractive?|test_interactive
Yes|Yes|No
|||
---|---|---|---
Tester|Hazen Babcock
Notes|(a), (b), (e), (g), (A), (L)
Date|2015-04-13
PLplot commit|09dee003
CMake version|2.8.12.2
Generator|\"Unix Makefiles\"
Platform|Lubuntu 14.04.2 LTS with system libraries
Pango/Cairo version|1.36.3/1.13.1
Qt version|Not Applicable
WxWidgets version|Not Applicable
Shared libraries?|Dynamic drivers?
Yes|Yes
Yes|No
No|No
CMake-based build tree?|test_noninteractive?|test_interactive?|ctest?
Yes|Yes|No|Yes
CMake-based installed examples?|test_noninteractive?|test_interactive
Yes|Yes|No
Traditional Installed examples?|test_noninteractive?|test_interactive
Yes|Yes|No
|||
---|---|---|---
Tester|Alan W. Irwin
Notes|(a), (b), (c), (d), (e), (f), (A), (K)
Date|2015-04-11
PLplot commit|99fe5f0
CMake version|3.0.2
Generator|\"MSYS Makefiles\"
Platform|MinGW-4.7.2/MSYS/Wine-1.6.1 with epa_built libraries
Pango/Cairo version|Not Applicable
Qt version|Not Applicable
WxWidgets version|Not Applicable
Shared libraries?|Dynamic drivers?
Yes|Yes
Yes|No
No|No
CMake-based build tree?|test_noninteractive?|test_interactive?|ctest?
Yes|Yes|Yes|Yes
CMake-based installed examples?|test_noninteractive?|test_interactive
Yes|Yes|Yes
Traditional Installed examples?|test_noninteractive?|test_interactive
Yes|Yes|Yes
|||
---|---|---|---
Tester|Alan W. Irwin
Notes|(a), (b), (c), (d), (e), (A)
Date|2015-03-29
PLplot commit|5f6e28f
CMake version|3.0.2
Generator|\"Unix Makefiles\"
Platform|Debian stable with epa_built libraries
Pango/Cairo version|1.35/1.12.14
Qt version|5.3.2
WxWidgets version|3.0.2
Shared libraries?|Dynamic drivers?
Yes|Yes
Yes|No
No|No
CMake-based build tree?|test_noninteractive?|test_interactive?|ctest?
Yes|Yes|Yes|Yes
CMake-based installed examples?|test_noninteractive?|test_interactive
Yes|Yes|Yes
Traditional Installed examples?|test_noninteractive?|test_interactive
Yes|Yes|Yes
Testing notes where lower case notes concern configuration and build options while upper case notes concern limitations on testing due to platform constraints ("not available"), arbitrary constraints imposed by the tester to simplify testing ("not installed" or "ignored"), or working around PLplot issues for the platform ("disabled").
* (a) Testing done on 64-bit (x86-64, also known as x64, x86_64, and AMD64, ) hardware.
* (b) Used parallel make option (-j4) for all builds, installs, and test targets.
* (c) Suitable dependent libraries have been installed on the system so there are no device drivers from the default list for this platform that are missing from this test.
* (d) Suitable compilers and bindings-related development packages have been installed on the system so there are no default bindings that are missing from this test.
* (e) Java, Python, Octave, Lua, and OCaml bindings/examples require shared PLplot libraries in order to work so were not available for testing for the static PLplot libraries case.
* (f) For this MinGW/MSYS/Wine platform test the build_plplot_lite epa_build target was used which simplifies testing by excluding the qt, cairo, and wxwidgets devices from tests. This was a \"lite\" test in other respects as well; i.e., the D, Java, OCaml, and Octave languages were not available on this platform.
* (g) This Lubuntu platform did not have Ada, Fortran, D, Java, or OCaml compilers installed and did not have development versions of Octave, Tcl/Tk, WxWidgets, Lua, shapelib, qhull, libLASi, or libharu installed. Therefore corresponding components of PLplot were disabled and not tested.
* (h) This MinGW/MSYS platform with no epa_built libraries has very few PLplot soft prerequisites available so tests on this platform necessarily only apply to a small number of the PLplot components (i.e., just the mem, ntk, null, ps, svg, wingcc, and xfig device drivers and just the c++, f95, and tcl bindings).
* (i) CMake was hand-built to make it depend on Qt4 rather than Qt5 to help avoid Qt4/Qt5 MSYS2 package conflicts.
* (j) g++ (but not gcc) -DUNICODE option required to make wxwidgets device driver buildable.
* (k) wxwidgets-only interactive tests using -DOLD_WXWIDGETS=ON to comprehensively test the old wxwidgets device driver.
* (A) No obvious configure, build, or install errors. No run-time errors in tests other than those noted in additional \"upper-case\" notes (if any).
* (B) Interactive tests disabled to avoid run-time errors.
* (B') Interactive tests ignored.
* (C) Traditional build of installed tests disabled to avoid build errors.
* (C') pkg-config not available therefore traditional build of installed examples not tested.
* (C'') pkg-config was installed, but traditional build had to be disabled because of runtime errors.
* (D) libqhull not available.
* (D') shapelib not available.
* (E) Swig not available therefore Java, Python, Octave, and Lua bindings not tested.
* (E') Swig was installed but Java, Octave, and Lua not installed so those bindings not tested.
* (E'') Java not available and Octave ignored.
* (E''') Java not available, Octave ignored, and Numpy not installed so java, octave, and python bindings not tested.
* (E'''') Java and Octave not available so java and octave bindings not tested.
* (F) Tk/Itcl/Itk not installed therefore not tested.
* (F') X11 not available. Therefore -dev tk and -dev tkwin not tested and Itcl/Itk not installed therefore not tested.
* (F'') X11 server was running, but was not picked up - DISPLAY variable? Therefore, dev tk and -dev tkwin not tested and Itcl/Itk not installed therefore not tested.
* (F''') X11, Itcl, and Itk not available. Therefore -dev xcairo, -dev tk, and -dev tkwin not tested and Itcl/Itk not tested.
* (G) Ada, D, and OCaml compilers not availables so Ada, D, and OCaml bindings not tested.
* (G') Ada disabled, D not available, and ocamlidl not available so Ada, D, and OCaml bindings not tested.
* (H) libharu, libLASi, wxwidgets not available so pdf, psttf, and wxwidgets device drivers not tested.
* (H') libharu not available so pdf device driver not tested.
* (H'') libharu and libLASi not available and wxwidgets ignored so pdf, psttf, and wxwidgets device drivers not tested.
* (H''') libharu and libLASi not available so pdf and psttf device drivers not tested.
* (I) libpango/libcairo not available therefore cairo devices not tested.
* (J) libQt not available therefore qt devices not tested.
* (K) Encountered intermittent (once only, repeat test did not have it) make.exe error: \"INTERNAL: Exiting with 1 jobserver tokens available; should be 4!\" I attribute this error to the problematic make.exe parallel-build option on MSYS. (See \u003Chttps://sourceforge.net/p/mingw/bugs/1950/> for other unsolved classical MSYS parallel build flakiness.) To work around this bug for the MinGW/MSYS platform should use the epa_build cmake option, -DNUMBER_PARALLEL_JOBS:STRING=1. However, this workaround should not be necessary for the Cygwin or MinGW-w64/MSYS2 platforms.
* (L) For this Lubuntu system with Qt4.8.6 system libraries, severe memory management issues (including segfaults) occurred for all attempts to use qt devices. Therefore, to complete this comprehensive test had to disable everything Qt related by using the script option
`--cmake_added_options \"DEFAULT_NO_QT_DEVICES=ON -DENABLE_qt=OFF\"`
* (M) gtk+-x11-2.0 development package not installed, therefore extXdrawable_demo test not run.
* (N) sip ignored so pyqt4 not tested.
DETAILED CHANGELOG FOR THIS RELEASE
commit 8943015acbffa0f2f7c0dfff85971ffde4ff5d51
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed Aug 12 09:46:27 2015 -0700
Update release notes to conform to what occurred for this release
M README.Release_Manager_Cookbook
commit 655ae2f51cd9231ddc24e09ad559b1b6ac8d9746
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed Aug 12 09:43:47 2015 -0700
Add source code for the D and Lua examples to our website
This change corrects a long-standing oversight that I just happened to
notice for this release.
M scripts/htdocs-gen_plot-examples.sh
M www/examples.php
commit 6a4c0888bf725dffd429a12c80368848fb3f7105
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Tue Aug 11 14:46:36 2015 -0700
5.11.1 preparation: Update all files that the check_all target indicates are inconsistent with other PLplot files
The check targets that are built as dependencies of the check_all
target typically generate a file in the build tree from some source
file and then compare that generated file with the corresponding file
in the source tree that was previously generated in exactly the same
way. If that comparison fails, then the usual step is to check the
differences, and if those seem reasonable (i.e., no mistake in the
source file that was used to generate the file) then cure the issue by
copying the latest generated file in the build tree on top of the
older generated file in the source tree. For the
bindings/swig-support/plplotcapi.i case, the comparisons are only
relevant to the one part of that file (the #defines) that is generated
so the cure in that case I edited
bindings/swig-support/plplotcapi.i to maintain consistency in the
Tested by Alan W. Irwin <airwin@users.sourceforge.net> on Linux using
both
make check_all
in the build tree and
scripts/comprehensive_test.sh --do_test_interactive no
in the source tree.
M bindings/f95/plplot_parameters.inc
M bindings/swig-support/plplotcapi.i
M bindings/tcl/plplot_parameters.h
commit 85bc37c04ddd5991052ccd3791753ba8c11f935f
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Tue Aug 11 13:50:03 2015 -0700
Update version and release date data for the forthcoming 5.11.1 release.
M cmake/modules/plplot_version.cmake
M doc/docbook/src/CMakeLists.txt
M www/examples.php
commit f148c6a0ea31959c7030952e5bd8a3234db61a90
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Tue Aug 11 13:48:42 2015 -0700
Tweak UTF-8 part of doc/docbook/README.developers.
M doc/docbook/README.developers
commit 4f49d208a5da3476d02a8ebe379b7b9be9e15b77
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Tue Aug 11 10:52:21 2015 -0700
Update of the release notes for the forthcoming 5.11.1 release
This update (which is likely the last update prior to the 5.11.1
release) includes a summary of all the improvements made during this
release cycle (based on looking at all the commit messages for this
release cycle). This update also adds a report of comprehensive test
success for the -DOLD_WXWIDGETS=ON case.
M README.release
commit f3f22e8be0880f225bc961d2750f4f40b7bcc2e1
Author: Tom Schoonjans <Tom.Schoonjans@bam.de>
Date: Mon Aug 10 15:56:17 2015 +0100
Removes quotation marks in plConfig.h.in
Removed for PLPLOT_VERSION_MAJOR, PLPLOT_VERSION_MINOR and
PLPLOT_VERSION_PATCH.
This is necessary to ensure that they are useful for comparing these
version numbers in preprocessor macros like:
\#if PLPLOT_VERSION_MAJOR >= 5 && \
PLPLOT_VERSION_MINOR >= 11
#define PLCALLBACK plcallback
\#else
#define PLCALLBACK plstream
\#endif
This patch ensures that the preceding code snippet compiles when using PLplot 5.11.0: right now it fails...
M include/plConfig.h.in
commit 67472567c195e3c0c1339c1e499efb3f9d84df3b
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Fri Aug 7 14:19:42 2015 -0700
Build system(s): make -DOLD_WXWIDGETS=ON work for all build systems and major configurations
Tested by: Alan W. Irwin <airwin@users.sourceforge.net> on Linux using
scripts/comprehensive_test.sh --cmake_added_options "-DDEFAULT_NO_DEVICES=ON -DPLD_wxwidgets=ON -DDEFAULT_NO_BINDINGS=ON -DENABLE_wxwidgets=ON -DOLD_WXWIDGETS=ON -DENABLE_cxx=ON" --do_test_noninteractive no --do_ctest no
and
scripts/comprehensive_test.sh --cmake_added_options "-DDEFAULT_NO_DEVICES=ON -DPLD_wxwidgets=ON -DDEFAULT_NO_BINDINGS=ON -DENABLE_wxwidgets=ON -DENABLE_cxx=ON" --do_test_noninteractive no --do_ctest no
M cmake/modules/wxwidgets.cmake
M examples/c++/CMakeLists.txt
M examples/c++/Makefile.examples.in
M examples/plplot_configure.cmake_installed_examples.in
M plplot_test/plplot-test-interactive.sh.in
commit a83e465186e3cc9c7c16236d8378f9d090cbadbf
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Fri Aug 7 13:25:31 2015 -0700
Comprehensive testing script: process cmake return code for installed examples
This fix insures if that if there is a cmake command failure for the
installed examples, the script immediately exits with appropriate
messages (just as already had been implemented when running the cmake
command for the core build).
M scripts/comprehensive_test.sh
commit 9653b7ce266623a0f9a9f4d5b356952afdcf24ec
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Thu Aug 6 02:14:11 2015 -0700
Update release notes for the forthcoming 5.11.1 release
These release notes contain descriptions of the more important changes
in ChangeLog.release_preliminary as well as tables (in SourceForge
markdown format) of all comprehensive testing done so far for this
release.
M README.release
commit 178b010930fd4b8f1cb874fe4e971a39ec7d7db7
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Thu Jul 30 13:54:08 2015 -0700
Propagate tcltk_in_plplot_library variable to the CMake-based build system for the installed examples
This change fixes a bug in the CMake-based build of the installed
examples when -DENABLE_DYNDRIVERS=OFF
M examples/plplot_configure.cmake_installed_examples.in
commit d64d9c6849e0971c1023584bc78fc7ab3a6a9289
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Thu Jul 30 08:23:11 2015 -0700
Revert "5867959 Fix to the multiple keypress bug on page advance" for cairo device driver
This is a temporary measure for the 5.11.1 release to return the cairo
device driver to essentially what existed for 5.11.0 to workaround the
following regression created by the above bug fix for the
test_extXdrawable target:
Scanning dependencies of target test_extXdrawable
The program 'extXdrawable_demo' received an X Window System error.
This probably reflects a bug in the program.
The error was 'BadWindow (invalid Window parameter)'.
(Details: serial 178 error_code 3 request_code 2 minor_code 0)
(Note to programmers: normally, X errors are reported asynchronously;
that is, you will receive the error a while after causing it.
To debug your program, run it with the --sync command line
option to change this behavior. You can then get a meaningful
backtrace from your debugger if you break on the gdk_x_error() function.)
make[3]: *** [examples/CMakeFiles/test_extXdrawable] Error 1
make[2]: *** [examples/CMakeFiles/test_extXdrawable.dir/all] Error 2
make[1]: *** [examples/CMakeFiles/test_extXdrawable.dir/rule] Error 2
make: *** [test_extXdrawable] Error 2
See further plplot-devel list discussion.
M drivers/cairo.c
commit ae54fcc46af39fa6d1798b898c07f7de8d8f52b0
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed Jul 29 15:00:26 2015 -0700
Comprehensive testing script: be selective in collecting environment variables
Instead of collecting all environment variables now collect only the
following relevant ones:
PATH
CC
CXX
FC
CMAKE_.*PATH (e.g., CMAKE_INCLUDE_PATH and CMAKE_LIBRARY_PATH)
.*FLAGS.* (e.g., CFLAGS, CXXFLAGS, and FFLAGS)
PKG_CONFIG_PATH
LD_LIBRARY_PATH
Being selective like this not only makes it easier to analyze the
results, but also should ease any privacy concerns there might have
been concerning the previous dumping of all environment variables set
for a user's computer into publicly distributed reports.
M scripts/comprehensive_test.sh
commit be85695721acac99135d7770e9fcd95da9bb70f6
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Mon Jul 27 10:00:36 2015 -0700
Build system: remove garbage from MINGWLIBPATH and BORLANDLIBPATH variables
These variables have been set with incorrect set command syntax
(i.e., with an extra DOCSTRING directive before the actual doc string)
from the very first. The result is the set command is interpreted as
specifying a CMake list as the value of the ordinary uncached
MINGWLIBPATH and BORLANDLIBPATH variables. That list for MINGWLIBPATH
is
<desired result>;CACHE;FILEPATH;DOCSTRING;"Path to MinGW import libraries"
and similarly for BORLANDLIBPATH.
These extraneous extra list values were than used as garbage HINTS to
help find wingcc libraries which did not affect the actual finding
since the <desired result> was also a HINT. So the net effect of this
bad set syntax was the libraries for wingcc were still found with a good
HINT plus a lot of garbage HINTS, but MINGWLIBPATH and BORLANDLIBPATH
are uncached, i.e., impossible for the user to modify.
These issues have now been addressed by removing the invalid extra
DOCSTRING directives in the set commands and the result is the garbage
hints are gone, and the results for MINGWLIBPATH and BORLANDLIBPATH
are now properly cached (as discovered with a mini test project on
Linux).
M CMakeLists.txt
commit 9b863855d326026bd0c7989bfbe614923472f59b
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Fri Jul 24 13:53:28 2015 -0700
Add account of all tests done on git repository that has been restored by SF staff after their "great outage".
A historical_repository_conversions/README_SourceForge_outage_restore
commit 4bbd72bd99e97046056a77da89fd09d915558887
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed Jul 22 18:45:32 2015 -0700
epa_build: Add valgrind configuration for the -DBUILD_THE_BUILD_TOOLS=ON case
Because the epa_build of valgrind and lapack_blas (another package
that is only configured for the -DBUILD_THE_BUILD_TOOLS=ON case) has
not been tried for the MSYS_PLATFORM=ON case, only configure these two
packages for now when BUILD_THE_BUILD_TOOLS=ON AND NOT
MSYS_PLATFORM=ON.
M cmake/epa_build/CMakeLists.txt
A cmake/epa_build/valgrind/CMakeLists.txt
commit add75b011b906440c24e6c55b069f2db699232b1
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sat Jul 18 20:57:08 2015 -0700
Traditional build system: make compiler options honor users' setting of compiler options and flags
For example, if the user specifies either
CC="gcc -g"
or
CC=gcc
CFLAGS="-g"
environment variables, the -g option will be used for the traditional
build of the C examples, and similarly for the C examples that
demonstrate Tk functionality and (for the equivalent environment
variables ADA and ADAFLAGS; CXX and CXXFLAGS; DC and DFLAGS; and FC
and FLAGS) the Ada, C++, D, and Fortran examples.
Tested by: Alan W. Irwin <airwin@users.sourceforge.net> on Linux
using
scripts/comprehensive_test.sh --do_ctest no --do_test_interactive no --do_test_build_tree no --do_test_install_tree no
M examples/c++/CMakeLists.txt
M examples/c/CMakeLists.txt
M examples/d/CMakeLists.txt
M examples/f95/CMakeLists.txt
M examples/java/CMakeLists.txt
M examples/tk/CMakeLists.txt
commit c531192daf861fed094bcb4345c94030d1d875af
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Fri Jul 17 14:25:40 2015 -0700
Handle user interrupts properly for the comprehensive testing script
This change insures collect_exit is called on user interrupt so the
report results up to the time of the interrupt will be collected in
the tarball.
M scripts/comprehensive_test.sh
commit 0a21ef1b835b42fbfbb2e76b2a41dda9c3ed51a4
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Fri Jul 17 11:05:12 2015 -0700
Build system, libcsironn, and libplplot: Revert back to using qhull/qhull_a.h #include form
As pointed out by Greg Jung, this #include style is recommended by
qhull_a.h to solve name clashes between the names of qhull headers
(e.g., io.h) and other system headers. As also pointed out by Greg
io.h is an important system header on Windows which resulted in
an actual io.h nameclash for him on Windows for a 32-bit MinGW-w64/MSYS2
comprehensive test of PLplot that he was attempting to do.
Note this reversion in #include style for qhull headers leaves Cygwin
out in the cold since the qhull packager there has gone out of his way
to change the name of the subdirectory where qhull headers are
installed from qhull to libqhull. I view that issue as a qhull
packaging bug for Cygwin so until that packaging bug is fixed, Cygwin
PLplot users will have to use a limited version of plgriddata without
qhull or else build their own qhull (which by default will install
headers in the correct qhull subdirectory rather than in the incorrect
libqhull subdirectory).
M cmake/modules/FindQHULL.cmake
M lib/nn/delaunay.c
M src/plgridd.c
commit 519c49490a6a7066f8cffe6f77948e2daef48052
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed Jul 15 14:12:58 2015 -0700
Improve comprehensive_test.sh script
The improvements are the following:
1. Make the summary of options state the options
with the same format users would need to specify them (and with quotes
wherever they might be required).
2. Make the script more robust by warning when no printenv command is
available, for that case removing corresponding stale file results, and
by checking that files exist before storing them in the tarball.
when there is no printenv command
available.
M scripts/comprehensive_test.sh
commit 44c7bfac24f83c04da896002e97b1c9d6e3df666
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed Jul 15 12:48:23 2015 -0700
epa_build: configure octave-lite
This is an epa_build of octave-4.0.0 that is configured to drop a
number of optional parts of octave.
As part of this change move the epa_build of lapack_blas from
the BUILD_THE_BUILD_TOOLS=OFF case to the BUILD_THE_BUILD_TOOLS=ON case
since the configuration of octave_lite (when
BUILD_THE_BUILD_TOOLS=OFF) requires that lapack_blas be built already.
Tested by: Alan W. Irwin <airwin@users.sourceforge.net> on Linux by
epa_building octave_lite and doing superficial run-time tests of
that result such as
octave:1> x=-pi:0.1:pi;
octave:2> plot(x,sin(x));
This test worked fine with default plotting (gnuplot), but segfaults
if a locally modified epa_build of plplot_lite is used that enables
octave and when the plotting is done with PLplot (using the
appropriate .octaverc such as the one configured in bindings/octave
for the PLplot build).
This same PLplot run-time test works fine when the octave binding of
PLplot is built against the system octave-3.6.2. I assume the
segfault for the octave-4.0.0 PLplot plotting case is caused by some
incompatibility between the octave binding of PLplot and octave-4.0.0,
rather than some issue with the epa_build of octave-4.0.0, but further
work will be needed to confirm that assumption.
M cmake/epa_build/CMakeLists.txt
A cmake/epa_build/octave_lite/CMakeLists.txt
commit 7760191850301bbe2f467f830855b528e7894de1
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed Jul 15 12:37:05 2015 -0700
epa_build: swig: update from patched 2.0.11 to patched 3.0.6
The present patch is to allow swig support for Octave-4.0.0. It
was derived by
wget https://github.com/swig/swig/pull/460.patch
mv 460.patch octave-4.0.0.patch
where https://github.com/swig/swig/pull/460 is Orion Poplawski's
pull request (which apparently has been accepted by swig developers)
to swig to fix this swig issue with octave-4.0.0.
Tested by: Alan W. Irwin <airwin@users.sourceforge.net> on Linux
with an epa_build test done by
make build_swig
with BUILD_THE_BUILD_TOOLS is ON
The result is not run-time tested yet.
M cmake/epa_build/swig/CMakeLists.txt
D cmake/epa_build/swig/octave-3.8.0.patch
A cmake/epa_build/swig/octave-4.0.0.patch
commit ecdf1574adb5e94fd3ed1f463a302d918021842a
Author: Arjen Markus <arjen.markus@deltares.nl>
Date: Mon Jul 13 12:53:18 2015 +0200
Add MSYS to the condition for including a .def file in the linking of the Fortran bindings on Windows.
Rationale:
If you use the MSYS shell, then CMake defines MSYS but not MINGW to identify theplatform, but the underlying compiler (gfortran) is the same. Hence to make sure that the right link information is used (in this case no .def file should be added to the link step), we need to cover the MSYS environment as well.
Tested on MingW-w64/MSYS2
M bindings/f95/CMakeLists.txt
commit ddfdb641012f7585f580a099f69c9b86f9937dc5
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Mon Jul 13 01:25:30 2015 -0700
Make comprehensive test script more verbose
M scripts/comprehensive_test.sh
commit ca6b96301a1aa4831112a6fcf0ff331bb4a0b4ea
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sun Jul 12 16:44:06 2015 -0700
epa_build: Configure an epa_build of lapack_blas
This build uses the CMake-based build system for the combination
of lapack and blas you get with
<http://www.netlib.org/lapack/lapack-3.5.0.tgz>.
Currently, no other epa_built package depends on this one, but
if it is eventually possible to epa_build octave, that package will
depend on lapack_blas.
Tested by: Alan W. Irwin <airwin@users.sourceforge.net> on Linux
by configuring epa_build and then running
make build_lapack_blas >& build_lapack_blas.out
ctest is automatically run as part of that build, and the ctest
results for that were
100% tests passed, 0 tests failed out of 86
M cmake/epa_build/CMakeLists.txt
A cmake/epa_build/lapack_blas/CMakeLists.txt
commit dd9f79e447263e219c65f7c4e8434625cd4c2de8
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sat Jul 11 15:57:32 2015 -0700
Build system: Bump CMake minimum version from 3.2.2 to 3.2.3
This change is motivated by what looks like an important fix in 3.2.3
(see <http://www.cmake.org/Bug/view.php?id=15546>) which works around
mingw32-make trailing backslash trouble which in turn (without the
workaround) interferes with the use of the "MinGW Makefiles" generator
for a number of different compilers.
The CMake minimum version remains at 3.0.2 for the Cygwin and Linux
platforms, i.e., they are unaffected by this change.
M CMakeLists.txt
M examples/CMakeLists.txt
commit 17f872b09055cbe7a532c05af4346d58d92c8de2
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sat Jul 11 10:07:01 2015 -0700
Comprehensive testing script: Create $prefix directory if it does not exist
This script bug was discovered as a result of comprehensive test results kindly provided by
Greg Jung.
M scripts/comprehensive_test.sh
commit d645c88fe0b243ccc9ed69507f5107769c1c4190
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sat Jul 11 00:47:40 2015 -0700
Build system: Propagate USE_DLL_SUBDIRECTORY=ON case correctly to octave binding
In all cases when USE_DLL_SUBDIRECTORY=ON, add an additional addpath
call for the dll subdirectory of the top-level build tree. This fix
is designed to solve the Cygwin octave issue recently discovered by
Arjen, but has not been tested on Cygwin yet.
During testing this fix on Linux I discovered that some
simplifications could be done, that is I was able to drop the trailing
":" from the octavedir environment variable list of directories to be
imported into octave for addpath to use, I did not need to set
LD_LIBRARY_PATH at all before I invoked octave, and I did not need to
push the PLPLOT_LIB environment variable from the
bindings/octave/.octaverc file that is configured as a convenience for
users (but which is not used in automated testing).
Tested by: Alan W. Irwin <airwin@users.sourceforge.net> on Linux
using
scripts/comprehensive_test.sh --cmake_added_options "-DDEFAULT_NO_DEVICES=ON -DPLD_ps=ON -DDEFAULT_NO_BINDINGS=ON -DENABLE_octave=ON" --do_test_interactive no
and also (to test the configured bindings/octave/.octaverc file)
make xwin
cd bindings/octave/
make plplot_octave
octave -q
x=-pi:0.1:pi;
plot(x,sin(x))
exit
M bindings/octave/octaverc.in
M plplot_test/plplot-test.sh.in
M plplot_test/test_octave.sh.in
commit 32be1b97820fcbd61730f9ebe9bea1effed87197
Author: Arjen Markus <arjen.markus@deltares.nl>
Date: Fri Jul 10 11:07:36 2015 +0200
Improve the logic for including the header file "unistd.h" in plfreetype. The original conditions led to a problem on MinGW/MSYS. As it turns that the main decision was to include this header file or not, the code now uses the macro that is specifically defined for this task.
Tested on MinGW/MSYS, but given the solution, this should work on all platforms, unless the controlling macro, PL_HAVE_UNISTD_H, is incorrectly defined there.
M src/plfreetype.c
commit 0f7231004787baf0cd08b20df6fdbef1bdcaede4
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed Jul 8 15:08:24 2015 -0700
Build system: Use more refined procedure for dealing with Octave-4 (Try 2)
Fix a bug in the previous commit on the Octave version to check against
(I used "3." rather than the correct "4.".)
M cmake/modules/octave.cmake
commit 2ad5aef13382f9918b94d2ccffb8ad73c3e7bcd2
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed Jul 8 13:57:51 2015 -0700
Build system: Use more refined procedure for dealing with Octave-4
If Octave-4 is detected then octave is automatically disabled with
a warning message unless the user specifically opts in by specifying
-DTRY_OCTAVE4=ON.
M cmake/modules/octave.cmake
commit 8cd0971dd17177d8baedcc03acdb5e1c6230167d
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed Jul 8 13:21:37 2015 -0700
Build system: Implement WARNING message for the Octave-4 case
Note that the WARNING message warns the users of potential build
errors but does not actually disable Octave in case someone wants to
experimentally try Octave-4 once the swig fix for Octave-4 #include
issues becomes available.
M cmake/modules/octave.cmake
commit 56b644e1e9be0d1c8b14a280a423bba7fcd01645
Author: Tom Schoonjans <schoon@users.sourceforge.net>
Date: Wed Jul 8 11:32:10 2015 -0400
Apply patch #31 from Tom Schoonjans. This patch allows the wincairo driver to compile and run successfully on 64-bit Windows.
M drivers/cairo.c
commit 6fb9dad4bfd104fbf2b156482f5f92eec469697f
Author: Hazen Babcock <hbabcock@fas.harvard.edu>
Date: Wed Jul 8 11:30:04 2015 -0400
Undo patch as I did not correctly specify the patch author.
M drivers/cairo.c
commit 6254e260e9a93d55fb8cc293847adb60c83219ed
Author: Hazen Babcock <hbabcock@fas.harvard.edu>
Date: Wed Jul 8 10:56:09 2015 -0400
Apply patch #31 from Tom Schoonjans. This patch allows the wincairo driver to compile and run successfully on 64-bit Windows.
M drivers/cairo.c
commit 273ec0dc38c1fe0288dad90cfeb51b07378dfa26
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed Jul 8 00:27:36 2015 -0700
Documentation: Fix file dependency of custom command that generates the man documentation
M doc/docbook/src/CMakeLists.txt
commit 4a8fb0a51ea5574a5a14932af1da961fd54bb5de
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed Jul 8 00:04:18 2015 -0700
Documentation: Fix leading blank issues in titles for DocBook-generated man and swig documentation pages
These issues are caused currently by extra blanks after ":" in the
title sections of the plcolbg, plcolbga, plscol0, plscol0a, and
plshade sections of doc/docbook/src/api.xml. However, these extra
blanks do not affect the normal DocBook output because of automatic
whitespace normalization. So instead of editing api.xml to temporarily
get rid of the issue until someone else inserts an extra leading blank
after a ":" in a title section, instead permanently solve the issue by
modifying the perl scripts used to generate the man pages and the swig
documentation to remove leading blanks from the generated title data.
Also, propagate the swig documentation that is generated (using
api2swigdoc.pl) by the check_swig_documentation target back to the
source tree using the usual procedure.
M bindings/swig-support/swig_documentation.i
M doc/docbook/bin/api2man.pl.in
M doc/docbook/bin/api2swigdoc.pl
commit 79310c1108ce68c873ce63ff6f9a80ffed60c3f6
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Tue Jul 7 22:59:28 2015 -0700
Swig-generated bindings documentation: Make that consistent with doc/docbook/src/api.xml
This is done by running
make check_swig_documentation
which generates from doc/docbook/src/api.xml documentation for the
swig-generated bindings and warns if that generated result is
different from the source-tree result. For this case where that
warning occurred, I diffed the two results to make sure the generated
version was OK, then copied that generated version back on top of the
source-tree version.
M bindings/swig-support/swig_documentation.i
commit e7b8f6305a70abcd0517d44abdfdfeef9202f424
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Tue Jul 7 17:22:34 2015 -0700
Build system: Fix another issue for a wxwidgets corner case and reinstate wxwidgets tests as part of comprehensive testing
The fix was for the traditional build where required cxx examples
builds were not done unless
ENABLE_cxx AND (PLD_ps OR ENABLE_qt)
The fix consisted of changing that condition to
ENABLE_cxx AND (PLD_ps OR ENABLE_qt OR ENABLE_wxwidgets)
which matters only for the corner case when both PLD_ps and ENABLE_qt
are false (as in the test below).
As discussed on list wxwidgets is still uncomfortably slow, e.g.,
the test_c_xcairo target requires
real 0m4.978s
user 0m2.848s
sys 0m0.368s
to finish while the test_c_wxwidgets target requires
real 0m59.189s
user 0m0.720s
sys 0m0.392s
to finish.
Since comprehensive testing does interactive testing 9 times (for the
build tree and two versions of the install tree for each of our 3
major configurations), the above length of time to finish the
test_c_wxwidgets test effectively adds 9 minutes to comprehensive
interactive testing. That amount of extra time required for
interactive testing is not too obnoxious for comprehensive testers
(although efficiency improvements to wxwidgets would be nice to
shorten that extra time considerably). Furthermore, the -np option
works now for wxwidgets which removes a major irritation with
comprehensive testing of wxwidgets. Therefore, I made the decision to
reinstate comprehensive testing of wxwidgets.
Tested by: Alan W. Irwin <airwin@users.sourceforge.net> on Linux by
running
scripts/comprehensive_test.sh --cmake_added_options "-DDEFAULT_NO_DEVICES=ON -DPLD_wxwidgets=ON -DDEFAULT_NO_BINDINGS=ON -DENABLE_wxwidgets=ON -DENABLE_cxx=ON" --do_test_noninteractive no --do_ctest no
M examples/CMakeLists.txt
M plplot_test/CMakeLists.txt
commit fc5b7d4b64c826fb97f6b4a50f595b165180f9da
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Tue Jul 7 16:59:16 2015 -0700
Build system: Fixes for various wxwidgets corner cases
Those corner cases are as follows:
1. libplplotwxwidgets depends on libplplotcxx so if ENABLE_cxx is OFF
must set ENABLE_wxwidgets to OFF as well
This corner case was discovered when attempting to use
scripts/comprehensive_test.sh --cmake_added_options "-DDEFAULT_NO_DEVICES=ON -DPLD_wxwidgets=ON -DDEFAULT_NO_BINDINGS=ON -DENABLE_wxwidgets=ON" --do_test_noninteractive no --do_ctest no
2. The wxwidgets device driver depends on wxPLViewer which depends on
libplplotwxwidgets so if ENABLE_wxwidgets is OFF, then the
PLD_<wxwidgets_device_names> should be set to OFF.
This corner case was discovered when attempting to use
scripts/comprehensive_test.sh --cmake_added_options "-DDEFAULT_NO_DEVICES=ON -DPLD_wxwidgets=ON -DDEFAULT_NO_BINDINGS=ON -DENABLE_wxwidgets=ON -DENABLE_cxx=ON" ----do_test_noninteractive no --do_ctest no
M bindings/wxwidgets/CMakeLists.txt
M cmake/modules/wxwidgets.cmake
M plplot_test/plplot-test-interactive.sh.in
commit ae0e9da925a871ac514087599e107c2ff1a9c424
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Mon Jul 6 16:05:14 2015 -0700
libplplot and libplplottcltk: make plplotLibDir symbol visible in more standard way
This change retains good test results on Linux (see below), and may
solve the current issue with plplotLibDir that has been discovered by
Arjen's recent comprehensive test on Cygwin.
Tested by: Alan W. Irwin <airwin@users.sourceforge.net> on Linux using
nm and ldd to check by hand for visibility issues for plplotLibDir within libplplot in
the nondynamic case and also by running
scripts/comprehensive_test.sh --cmake_added_options "-DDEFAULT_NO_DEVICES=ON -DPLD_tk=ON -DPLD_xwin=ON -DPLD_ntk=ON -DPLD_tkwin=ON -DPLD_ps=ON -DDEFAULT_NO_BINDINGS=ON -DENABLE_tcl=ON -DENABLE_itcl=ON -DENABLE_tk=ON -DENABLE_itk=ON" --do_test_interactive no
to mimic Arjen's recent comprehensive test on Cygwin (but for speed
just focussing on the Tcl/Tk part of that comprehensive test that
would be affected by plplotLibDir symbol visibility issues).
M bindings/tcl/tclAPI.c
commit fc34dee92457bdcfab3101b7f3cad3d173ac826c
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sat Jul 4 14:56:59 2015 -0700
Comprehensive testing script: Output git commit id of the PLplot version being tested
M scripts/comprehensive_test.sh
commit 59b9b268d8173b4f621c5f3c1705462bb7c4f3cd
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sat Jul 4 09:25:30 2015 -0700
Style previous commits
M utils/wxplframe.cpp
M utils/wxplviewer.cpp
commit 27581602c7eec6a5a853ee1d3e915e54a88ade10
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Sat Jul 4 11:43:14 2015 +0100
removed debug code accidentally left in from last commit
M drivers/wxwidgets_dev.cpp
commit 48a3b5fb5e9264489592eb9c383c05bafdc0e9e5
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Sat Jul 4 11:34:31 2015 +0100
Removed exit() calls from wxPLViewer
There were two exit calls in wxplframe.cpp. Calling exit bypasses all
the cleanup code including freeing any system resources that have been
allocated. The exit call in the constructor of wxPlFrame has been
replaced with a throw, which is caught in wxPlViewerApp::OnInit, which
then returns false causing the application to close with an error code.
The exit call IN wxPlFrame::ReadTransmission has been replaced by a call
to wxWindow::Close() which causes the frame to close gracefully and
because it is the only top level window this causes the application to
close too.
M drivers/wxwidgets_dev.cpp
M utils/wxplframe.cpp
M utils/wxplviewer.cpp
commit 5e74b6a6403bd2c4a38d82c4dea574fd7c39811c
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Fri Jul 3 12:43:33 2015 +0100
Modification to previous wxPLViewer optimisation
Added multiple retries with a short delay between each in checking for
more transmitted data. This was because requesting text size took
slightly longer than one extra check, but was much quicker than waiting
for the next timer event.
M utils/wxplframe.cpp
commit 0a13b1a822205134d9a174a3ab3ce410745f8740
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Fri Jul 3 12:16:42 2015 +0100
Optimisation of wxPLViewer
We now check for new commands multiple times each time we get a timer
event. The reading of the transmission data has been moved to a separate
funtion which returns false if it checked the buffer and found there was
nothing to read and returns true otherwise. When we receive a timer
event we repeatedly call this function until we get a false. On Windows
with wxWidgets 3.0 the speed up is not huge, but on CentOS with
wxWidgets 2.8 there is a ten times speed increase for example 8. The
difference is presumably to do with the different timers and event
handling between the two systems.
M utils/wxplframe.cpp
M utils/wxplframe.h
commit 3c4facf97d6a1fcec1881e7848ab8f5cc7390a7d
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Fri Jul 3 12:16:17 2015 +0100
Style previous commits
M drivers/wxwidgets_comms.h
M drivers/wxwidgets_dev.cpp
M utils/wxplframe.cpp
M utils/wxplframe.h
commit 2476c06188136aa169e5cf296acbed058313d8de
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Fri Jul 3 10:18:58 2015 +0100
Changed the way fonts are passed to wxPLViewer and created when received
This fixes a but when using Ubuntu with wxWidgets 3.0 and hopefully
gives a speed increase.
When the console mode part of the app created a wxFont to get a text
description to pass to wxPLViewer, this caused a crash on the above
system. Instead now we simply convert the plplot native font parameters
to those required for a wxFont and pass these instead. When the
wxPLViewer receives the parameters it uses wxTheFontList to create the
font. This caches fonts so speeds up multiple calls.
M drivers/wxwidgets_comms.h
M drivers/wxwidgets_dev.cpp
M utils/wxplframe.cpp
commit aeac20cf9bcc4f9058e96da2bb86d7d7de19a0ff
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Wed Jun 24 12:47:20 2015 +0100
Optimisation of wxPLViewer to speed up plotting
We were getting very slow rendering which proved to be related to requesting
text size from the viewer. To improve this I removed a mutex from
wxDevice::DrawTextSection as it was not needed and now do not immediately slow
the the buffer poling in wxPLViewer if it finds no data.
Signed-off-by: Phil Rosenberg <p.d.rosenberg@gmail.com>
M drivers/wxwidgets_dev.cpp
M utils/wxplframe.cpp
M utils/wxplframe.h
commit 9c43f62e4b37a3395d977e5801358d3ac9472a20
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Wed Jun 24 10:55:22 2015 +0100
Optimisation for the wxWidgets driver - finding wxPLViewer
Previously the wxWidgets driver enumerated the entire build directory when
hunting for wxPLViewer which took 75% of the execution time. Now it just
checks the utils directory.
Signed-off-by: Phil Rosenberg <p.d.rosenberg@gmail.com>
M drivers/wxwidgets_dev.cpp
commit 8e6c5bfe594753166fae56a49410dd223e19ce6e
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Tue Jun 23 08:32:10 2015 -0700
Drop execute permission bits from files that appear not to need them
I used the command
find . -type f -a '!' -name "*.sh*" -a '!' -name "*.pl*" \
-a '!' -perm u=rw,go=r | \
grep -Ev '(.git|debian|examples/tcl|examples/python)' |less
to find files that were potential candidates for wrong permissions.
M bindings/octave/README
M bindings/tk/plcolor.tcl
M data/cglobe.map
M data/globe.map
M data/plstnd5.fnt
M data/plxtnd5.fnt
M data/usa.map
M data/usaglobe.map
M src/CMakeLists.txt
commit 8025ebeac568505dd67688aa46b2a56c527fbc11
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Mon Jun 22 09:39:56 2015 -0700
Style previous commits
M drivers/cairo.c
M drivers/qt.cpp
M drivers/tkwin.c
M drivers/wingcc.c
M drivers/xwin.c
M include/plplotP.h
M src/plcore.c
M src/plpage.c
commit 78344dfc901e8f66c1468055fe33423f4fa5efff
Author: Jim Dishaw <jamesd47@users.sourceforge.net>
Date: Sun Jun 21 22:43:00 2015 -0400
Corrections to the cairo.c file to fix compilation error
- When cairo.c was modified for the multiple keypress bug, the local
variables need in the wait function were not defined.
- The required variables were moved from the EOP handler, where they
were no longer needed, to the wait function handler.
M drivers/cairo.c
commit 009fa6b4badb5922b2e306e02ab4603342616803
Author: Jim Dishaw <jim@dishaw.org>
Date: Mon Jun 15 21:51:54 2015 -0400
Updated the wingcc driver to reflect the changes for the keypress bug
M drivers/wingcc.c
commit 5867959e87e64e08ed2ed57bc66a3cfa6e22450f
Author: Jim Dishaw <jim@dishaw.org>
Date: Mon Jun 15 12:10:04 2015 -0400
Fix to the multiple keypress bug on page advance
- The "wait for user input" is not part of an EOP.
-- A new function plP_wait() was created
-- plP_wait() calls were added next to plP_eop() calls to generate
a wait (if specified by nopause) after an EOP event
- The plot buffer was modified to insert an EOP into the buffer
-- No wait is inserted into the plot buffer because the plot buffer
only regenerates plots
- A new entry was added to the PLDispatchTable for the wait function
-- The table initialization sets all function pointers to NULL via
a memset
-- If a driver does not specify a wait function, either it is not
needed (e.g. the ps driver) or it is part of the EOP handler
(legacy behavior)
- The following drivers were modified to support the new implementation: cairo, qt, tkwin, and xwin. The wingcc update will be in a seperate fix.
M drivers/cairo.c
M drivers/qt.cpp
M drivers/tkwin.c
M drivers/xwin.c
M include/disptab.h
M include/plplotP.h
M src/plbuf.c
M src/plcore.c
M src/plpage.c
commit 602bb49d7889a50ce0f03a797d8e19e966703224
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sun Jun 21 17:13:10 2015 -0700
Driver documentation: use correct names of files that need updating for new device driver
M drivers/README.drivers
commit 87eed4ba0dbe18ff8a699fa3a7143ee47eb53736
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sun Jun 21 17:04:47 2015 -0700
Drivers: Purge long-retired device drivers from the source tree
The purged drivers are dg300, hpgl, impress, linuxvga, ljii, ljiip,
pbm, and tek.
The purged information does not include the DocBook documentation of
our device drivers which needs a complete rewrite rather than simply
removing any mention of these device drivers.
Tested by Alan W. Irwin <airwin@users.sourceforge.net> on Linux using
scripts/comprehensive_test.sh --do_test_interactive no
M bindings/tk-x-plat/Plplotwin.tcl
M cmake/modules/drivers-init.cmake
M cmake/modules/drivers.cmake
D cmake/modules/linuxvga.cmake
D drivers/dg300.c
D drivers/dg300.driver_info.in
D drivers/hpgl.c
D drivers/hpgl.driver_info.in
D drivers/impress.c
D drivers/impress.driver_info.in
D drivers/linuxvga.c
D drivers/linuxvga.driver_info.in
D drivers/ljii.c
D drivers/ljii.driver_info.in
D drivers/ljiip.c
D drivers/ljiip.driver_info.in
M drivers/mem.c
D drivers/pbm.c
D drivers/pbm.driver_info.in
D drivers/tek.c
D drivers/tek.driver_info.in
M examples/CMakeLists.txt
M examples/plplot_configure.cmake_installed_examples.in
M include/drivers.h
M include/plDevs.h.in
M include/plcore.h
M include/pldll.h.in
M include/plplotP.h
M plplot_test/plplot-test.sh.in
M src/plargs.c
commit 2db68f634b5a4087fda87b73a5cf4e6109c1ca01
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Sat Jun 20 13:32:29 2015 +0100
Impliment use of nopause in the wxWidgets driver
Added a chek for nopause for the wxwidgets driver when using wxViewer.
This was a missing feature to now. To do this the device sends a new
transmissionEndOfPageNoPause ratehr than a transmissionEndOfPage
command. This causes the viewer to immediately display the transmitted
page rather than waiting for user input.
In addition a transmissionClose has been added which causes the viewer
to exit. if pls->nopause is set then this command will be sent just
before the device destructor is called and it will cause the viewer to
close.
M drivers/wxwidgets.cpp
M drivers/wxwidgets.h
M drivers/wxwidgets_comms.h
M drivers/wxwidgets_dev.cpp
M utils/wxplframe.cpp
commit 63ba9005351d385b0ba4d56f1931c78c845b8d50
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Sat Jun 20 12:14:54 2015 +0100
changes a sizeof for consistenct in wxwidgets_dev.cpp
in wxPLDevice::TransmitBuffer made sure we used the transmissionLocate
variable in sizeof calls when sending a transmissionLocate command. This
is just for consistency as all command variables have the same size.
M drivers/wxwidgets_dev.cpp
commit 73e872a8deded50b2135fb881f2860e195d78c09
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Thu Jun 18 19:07:32 2015 -0700
Spelling: concatanate ==> concatenate
This change only affects commentary and documentation.
M README.Release_Manager_Cookbook
M cmake/modules/drivers.cmake
commit 12f1c938950edc1e891865a37e1e002b6bc7adf7
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed Jun 17 19:10:48 2015 -0700
Build system: fix logic bugs that were revealed by the -DPLD_tk=OFF -DPLD_tkwin=OFF case and the -DENABLE_itk=OFF case
1. Change
NOT ENABLE_DYNDRIVERS ==> tcltk_in_plplot_library
everywhere as a test whether the tclmatrix and plplottcltk code
is compiled as part of the plplot library or whether the tclmatrix
and plplottcltk libraries are actually built and linked to separately.
The tcltk_in_plplot_library variable is equivalent to
NOT ENABLE_DYNDRIVERS AND (PLD_tk OR PLD_tkwin)
and the effective addition of that AND (PLD_tk OR PLD_tkwin) is the real
fix here.
This change fixes at least an undefined symbol issue with the
plplottcltk_Main library that shows up for the -DPLD_tk=OFF
-DPLD_tkwin=OFF case, but I suspect there are other corner cases that
are fixed by this change as well.
2. Change to setting the -DUSINGDLL compile flag on library source
code so that occurs just before when the library is built to make sure
that flag is used for all source code for a given library. This
change at minimum fixes a visibility bug in the build of the
plplottcltk_Main library that showed up when -DPLD_tk=OFF
-DPLD_tkwin=OFF, but I suspect there are other corner cases that are
fixed by this change as well.
3. Test PLD_tk rather than ENABLE_tk when deciding
whether to configure bindings/tk/pkgIndex.tcl to load the TEA-based
Pltk package. That package is based exactly on the tk device driver
code so this fix solves a bug where CMake errors out at generation
time with the message
CMake Error at cmake/modules/plplot_functions.cmake:251 (file):
Error evaluating generator expression:
$<TARGET_FILE_DIR:tk>
No target "tk"
[...]
for the corner case where PLD_tk=OFF and ENABLE_tk=ON
4. Fix several issues with whether various Tcl/Tk/Itcl/Itk tests are
configured or not.
4a. test_tk_0[1-4], test_tk_plgrid, and test_plserver_standard_examples
no longer incorrectly depend on PLD_tk.
4b. test_plserver_runAllDemos no longer incorrectly depends on PLD_tk and
ENABLE_itkX.
4c. test_wish_runAllDemos now correctly depends on PLD_tkwin and no longer
incorrectly depends on PLD_tk and ENABLE_itkX.
Tested by: Alan W. Irwin <airwin@users.sourceforge.net> on Linux using
scripts/comprehensive_test.sh --cmake_added_options "-DDEFAULT_NO_DEVICES=ON -DPLD_tk=OFF -DPLD_xwin=ON -DPLD_ntk=ON -DPLD_tkwin=OFF -DPLD_ps=ON -DDEFAULT_NO_BINDINGS=ON -DENABLE_tcl=ON -DENABLE_itcl=ON -DENABLE_tk=ON -DENABLE_itk=ON"
N.B. note the -DPLD_tk=OFF -DPLD_tkwin=OFF options in the above command.
In addition, the test_plserver_runAllDemos and test_wish_runAllDemos
test targets (which are currently excluded from comprehensive testing
because of some issues with runAllDemos.tcl) were built for the
-DENABLE_itk=OFF case proving those tests could be run (as well as
usual) independent of Itk.
M bindings/CMakeLists.txt
M bindings/tcl/CMakeLists.txt
M bindings/tk-x-plat/CMakeLists.txt
M bindings/tk/CMakeLists.txt
M cmake/modules/tk.cmake
M examples/CMakeLists.txt
M examples/tk/CMakeLists.txt
M examples/tk/Makefile.examples.in
commit d07f1e2907a75dc964cd8825c60aa6d4bc9324b3
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed Jun 17 19:05:12 2015 -0700
Build system: Only calculate and print out TKLIB_COMPILE_FLAGS when actually needed
This change is primarily motivated by removing confusing and unneeded
Tk output information when ENABLE_tk=OFF.
M cmake/modules/tk.cmake
commit 1137ca9e6b76f9b81964068fccc85aae0d363e1b
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed Jun 17 14:14:05 2015 -0700
Build system: Fix potential inconsistencies in itcl and itk variables
With the old CMake logic it was possible in a certain corner case
(found by one of Arjen's tests when an iwidgets package was not
installed on Cytwin) for HAVE_ITCL, HAVE_ITCLDECLS_H to be set
inconsistently with ENABLE_itcl and HAVE_ITK inconsistently with
ENABLE_itk. The result was a build error with missing symbols which I
verified also on Linux. This fix makes such inconsistencies
impossible, and therefore should make it possible to build PLplot when
iwidgets is not found.
Tested by Alan W. Irwin <airwin@users.sourceforge.net> on Linux
by temporarily looking for iwidgetsxx rather than iwidgets to
emulate a missing iwidgets. The result was bad
make plplottcltk
ldd -r bindings/tcl/libplplottcltk_Main.so
results before the fix, and good results afterwards.
M cmake/modules/tcl-related.cmake
commit e5c84166042942c15d23464a8cd22c5fd3cc37bc
Author: Jim Dishaw <jim@dishaw.org>
Date: Wed Jun 17 10:43:34 2015 -0400
Corrects an off by 1 error in plbuf.c identified by Greg Jung
- In wr_command() and wr_data() the check_buffer_size() was called with
size of the data being written without considering the two-byte alignment
padding.
- On odd data sizes (e.g. 1 byte), the check_buffer_size() call would be
one byte smaller than what is required to maintain alignment.
- This could result in unpredictable memory access boundary errors.
M src/plbuf.c
commit de4e65fffbb87f0eb0bcac48ec75920e1a57481e
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed Jun 17 13:20:03 2015 -0700
Developer documentation: add reference to advice on improving both git commit messages and commits
M README.developers
commit f4db3e8162d7ffde32bb2baebe34c679d9c80761
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Tue Jun 16 22:42:29 2015 -0700
Build system: Avoid redundant linking of tclmatrix and plplottcltk libraries
If ENABLE_DYNDRIVERS=OFF then all code in both the tclmatrix and
plplottcltk libraries appears in the plplot library. So for this case
we replace redundant linking to the tclmatrix, plplottcltk, and plplot
libraries by just linking to the plplot library. In fact, for this
case we don't bother to build or install the tclmatrix and plplottcltk
libraries at all.
Note, the working hypothesis behind these changes is redundant linking
might cause memory management issues on library exit. At least that is
our experience with a similar redundant linking problem (since fixed)
for the plplotqt library. But regardless of that potential issue, I
also feel it is cleaner not to use redundant linking.
In addition to the above fundamental and intrusive change, we have
taken this opportunity to also do the following:
1. Rewrote bindings/tk/pkgIndex.tcl.in and
bindings/tk-x-plat/pkgIndex.tcl.in to be exactly the same as
bindings/tcl/pkgIndex.tcl.in aside from the name of the TEA-based
package being loaded. The corresponding CMakeLists.txt files have
been changed to configure these files correctly in all three cases.
This approach works on Unix (which fixes a bug where the Pltk and
PLplotter packages would not load for the ENABLE_DYNDRIVERS=OFF case
before) and should also work on Windows, but that latter platform
remains to be tested.
2. Updated instructions in examples/tcl/README.tcldemos and
examples/tk/README.tkdemos about running the Tcl/Tk/Itcl/Itk
examples by hand (especially useful for valgrind runs).
3. I don't think it is an improvement due to dropping redundant linking
but for examples/tk/wish_standard_examples(.in) it appears examples
2 and 24 now work without memory management issues or any other kind
of run-time error and example 20 works also without nopause issues.
Therefore, I have added 2, 20, and 24 to the examples that are run
by this script.
4. Add test_pltcl_standard_examples to the dependencies of test_tcl
and test_interactive targets for the Windows case. I have made this
change because we have good memory management results for
test_pltcl_standard_examples on the Linux platform according to
valgrind, and Arjen's previous test of this target where he reported
memory management issues on Cygwin may be an artifact of his previous
inconsistent Tcl/Tk/Itcl/Itk installation on that platform.
5. Sort out visibility issues for the tclmatrix and plplottcltk code
that is included in the plplot library when ENABLE_DYNDRIVERS=OFF.
Tested by: Alan W. Irwin <airwin@users.sourceforge.net> on Linux
using the command
scripts/comprehensive_test.sh --cmake_added_options "-DDEFAULT_NO_DEVICES=ON -DPLD_tk=ON -DPLD_xwin=ON -DPLD_ntk=ON -DPLD_tkwin=ON -DPLD_ps=ON -DDEFAULT_NO_BINDINGS=ON -DENABLE_tcl=ON -DENABLE_itcl=ON -DENABLE_tk=ON -DENABLE_itk=ON"
I also tested the tclsh and wish examples in
examples/tcl/README.tcldemos and examples/tk/README.tkdemos using
valgrind. For some still unknown reason (not related to redundant
linking which is now gone) there continues to be severe memory
management issue when attempting to exit wish after running the
standard examples under wish. All other results including those from
examples 2, and 24 were valgrind clean (until the end) which is why I
now include example 2 and 24 in the test, see comment 3 above.
M bindings/python/CMakeLists.txt
M bindings/tcl/CMakeLists.txt
M bindings/tcl/pkgIndex.tcl.in
M bindings/tk-x-plat/CMakeLists.txt
M bindings/tk-x-plat/pkgIndex.tcl.in
M bindings/tk/CMakeLists.txt
M bindings/tk/pkgIndex.tcl.in
M examples/CMakeLists.txt
M examples/tcl/CMakeLists.txt
M examples/tcl/README.tcldemos
M examples/tk/CMakeLists.txt
M examples/tk/Makefile.examples.in
M examples/tk/README.tkdemos
M examples/tk/wish_standard_examples.in
M include/pldll.h.in
M src/CMakeLists.txt
commit 73fd9a4623eb1bf7b2564168b1dd8cfe75da58ab
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Tue Jun 16 09:39:01 2015 -0700
Build system: Fix "spaced" pathname bug for concatenate_pkgIndex.tcl target (addendum)
The first attempt to fix this issue only quoted the pathnames of the
files to be concatanated. This addition to that fix also quotes the
pathname of the resulting concatanated file.
M bindings/CMakeLists.txt
commit bed8293351198be75311fa09258a954d68b849d9
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Mon Jun 15 16:06:59 2015 +0100
Restyling only commit
M drivers/wxwidgets.h
M drivers/wxwidgets_dev.cpp
M src/plbuf.c
M src/plcore.c
commit 32c5cf87fb484595f08ebb95983e375d8daeead1
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Mon Jun 15 09:33:39 2015 -0700
Build system: Fix "spaced" pathname bug for concatenate_pkgIndex.tcl target
M bindings/CMakeLists.txt
commit 6cde620d2c262dac3e1861dd9414d7b1b7f68c46
Author: Jim Dishaw <jim@dishaw.org>
Date: Wed Jun 3 14:21:40 2015 -0400
Fixes to correct builds on WIN32
M src/plbuf.c
M src/plfreetype.c
M src/plmetafile.c
M src/plstdio.c
commit 52462b91eebe41e1695deaebc39dc2b377d48148
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Thu Jun 11 16:58:00 2015 +0100
Replaced adding clipping region to every text and line command (which
could end up with a lot or redundant buffer in the data) with a
plbuf_clip function which adds the clip data to the buffer when plP_sclp
is called.
M include/drivers.h
M include/metadefs.h
M src/plbuf.c
M src/plcore.c
commit a59c46ef5a02b48b77ba0ce818a4e052a5954748
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Tue Jun 9 18:24:15 2015 +0100
Added clipping region to the buffer when we draw a line. this matches
what is done for text and ensures that lines with a width bigger than 1
pixel do not spill over the viewport boundaries.
M src/plbuf.c
commit 74ce8cf58d0bfd1d2669ed487bec9ad06975f2cf
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Fri Mar 6 12:09:21 2015 +0000
Added clipping limits to lines in wxWidgets so that fat lines do not spill out over the plot
edges. Still needs some testing. Doesn't work in PLViewer as clip limits do not seem to be saved
in the buffer. Not entirely sure where/when clip limits are set so this should be checked.
M drivers/wxwidgets.h
M drivers/wxwidgets_dev.cpp
commit 868f7903fde82771e041d1fdcb42a9044fd07905
Author: Jim Dishaw <jim@dishaw.org>
Date: Mon Jun 8 18:16:43 2015 -0400
Fixed a NUL termination bug in plbuf.c
When storing non-unicode strings, the NUL character was not being stored in
the buffer. This causes problems with drivers that use non-unicode strings
when the plot buffer is replayed.
M src/plbuf.c
commit 354bffe5d676f4134f59caead68fe8a76832963e
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sat Jun 6 10:42:07 2015 -0700
Build system: use PLD_wxwidgets=ON by default
Recent changes have cured the build trouble on Linux for CMake-2.8.12 so
set PLD_wxwidgets=ON by default again.
M cmake/modules/drivers-init.cmake
commit 215a6f6fd43c0c901ffd9836b873da864bbf2eed
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sat Jun 6 10:06:33 2015 -0700
Build system: FindwxWidgets.cmake: order of search for aliases of wx-config
The usual CMake philosophy is to search first for most specific forms
of names in descending order by version then generic. So I have
followed that philosophy for the possible aliases of wx-config. I
also included 2.9 in the range of versions from 3.0 to 2.8 because
that is also a possibility. Finally, I replaced the previous logic of
separate find_program calls by using the NAMES signature for
find_program.
M cmake/modules/FindwxWidgets.cmake
commit 1d12817a45734da20d2ca634840568f6a82b6e35
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Sat Jun 6 13:46:46 2015 +0100
Check for wx-config-3.0 and wx-config-2.8 in FindwxWidgets.cmake. These
executables or used on Cygwin instead of wx-config
M cmake/modules/FindwxWidgets.cmake
commit c4951f9a7c788bcf56a5d60e7b961339a90514de
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Sat Jun 6 11:30:07 2015 +0100
Added limits header to wxwidgets_dev.cpp. It seems that on windows this
was included circuitously, but not on linux meaning that use of
std::limits caused a compile failure.
M drivers/wxwidgets_dev.cpp
commit bacd63824ed69f9276767fb164cd748795f9cefd
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sat Jun 6 00:34:06 2015 -0700
Build system: emergency temporary response to current wxwidgets build issues
I have temporarily disabled the wxwidgets device and library because
I cannot get those to build on Linux when the wxwidgets sofware version is
either 2.8.12 or 3.0.2.
Until these build issues are addressed, those who want to test
wxwidgets builds on any platform will have to specify
-DPLD_wxwidgets=ON.
M cmake/modules/drivers-init.cmake
commit 3bd00d6cfce4c6f1090662eb28b38d0989bd57d8
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sat Jun 6 00:32:47 2015 -0700
Build system: implement wxwidgets version check for 2.8.12 or higher
This could easily be modified to 3.0.0 or higher if necessary.
M cmake/modules/wxwidgets.cmake
commit 63bdf93b5f499ea025e8f13ba51d52c2aacf1d46
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Fri Jun 5 11:27:51 2015 -0700
Build system: bug fix for plplotada library naming convention
For some reason (not yet analyzed but my guess is the problem occurs
for every Windows platform where UNIX is not true, i.e., all other
platforms than Cygwin) the naming convention for the plplotada library
on MinGW is not consistent with our other libraries so there has been
a long-standing workaround for that issue.
That workaround errors out on Cygwin because on that platform a very
different naming convention is used. Furthermore, it appears the
naming convention for the plplotada library on Cygwin is consistent
(i.e., no naming-convention workaround of any kind is needed for that
platform).
Thus, this bug fix simply drops the naming-convention workaround
for the Cygwin case. But for all other Windows platforms a
more fundamental fix is required than the current workaround.
M bindings/ada/CMakeLists.txt
commit 04ba6d679fe447ef376b804c75b017648f992079
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Fri Jun 5 14:29:42 2015 +0100
Added an append buffer escape and a flush remaining buffer escape and an
AppendBuffer function to wxPLplotstream. The use is pretty obvious in
that PLESC_APPEND_BUFFER appends data to the end of the current buffer
and PLESC_FLUSH_REMAINING_BUFFER flushes any buffer that hasn't been
rendered. A slight restructuring of the plRemakePlot() function in
plbuf.c allowed this.
The impact is a massive speed increase of wxPLViewer when using lots of
flusheses, for example the stripchart in example 17 ground to a halt in
the previous version which used PLESC_IMPORT_BUFFER and plreplot to
replot the whole buffer after each flush.
Also added an esc_old variable in plbuf_control. This is just useful for
debugging buffer coruption issues and isn't actually used by the library
(so will probably get optimised away anyway in release builds)
M bindings/wxwidgets/wxPLplotstream.cpp
M bindings/wxwidgets/wxPLplotstream.h
M include/drivers.h
M include/plplot.h
M src/plbuf.c
M utils/wxplframe.cpp
commit b556d4831e28b87068867584f3e9c872a2d6dc68
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Thu Jun 4 22:25:12 2015 +0100
Added the following optimisations to text (and hence point) drawing in
the wxWidgets driver.
Created a fontgrabber clas which is used to create a wxFont. It also
remembers the previous font which can be reused if the same font is
requested repeatedly (as is generally the case when drawing points). It
turned out that in profiling it turned out that around half the
rendering time seemed to be spent generating fonts. The GetFontMethod of
the device has been removed.
Store the previous string width if it is just a single character so that
if the same character with the same font is rendered repeatedly (as is
generally the case when drawing points) then the size can be reused.
This avoids sending text size requests to wxPLViewer and waiting for
responses.
Tidied up some unused text variables from the dev class including moving
m_underlined to a local variable/function argument in the text methods.
M drivers/wxwidgets.h
M drivers/wxwidgets_dev.cpp
commit 9423d1542c45f1809ee850db6b079c6fb1b5cb11
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Wed Jun 3 19:04:41 2015 +0100
Added setting the subpage and setting the number of subpages to the
buffer. This is required in order to correctly do a clear. Setting the
number of subpages is performed in plbuf_bop as well in order to catch
the initial state.
M drivers/wxwidgets_dev.cpp
M include/drivers.h
M include/metadefs.h
M src/plbuf.c
M src/plcore.c
M src/plpage.c
commit e8cb5f839320683de70b5eb2f971c402c73e95f0
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Thu Jun 4 03:34:32 2015 -0700
Build system: Work around linking issue for traditional static build
One minor bug fix to previous fix for this issue. On some platforms,
e.g., Mac OS X, CMAKE_CXX_IMPLICIT_LINK_DIRECTORIES can contain not
only the usual short CMake-style library names but also the full path
name for a library. So implement an if(EXISTS...) check for this
possibility if find_library cannot find the library based on the
asumption that the CMAKE_CXX_IMPLICIT_LINK_DIRECTORIES element is a
short CMake-style library name.
M cmake/modules/pkg-config.cmake
commit 4e99275c4fe9b49c917d40c924aa50cbc0df7fa5
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed Jun 3 12:40:05 2015 -0700
Build system: wxwidgets: more verbose reporting of wxwidgets find results
M cmake/modules/wxwidgets.cmake
commit aa4bc6b9e494df658f01b2a4768bf33a49ce3465
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed Jun 3 12:32:29 2015 -0700
Build system: wxwidgets; Adopt modified version of latest find module
For this commit, the cmake/modules/FindwxWidgets.cmake file was copied from CMake-3.2.1
and modified appropriately in the licensing section and also the
include(FindPackageHandleStandardArgs) command so we can use this file
as part of PLplot. Also, update this file with the WX_LIB_DIR_PREFIX
portion of Phil's previous patch. (The other part of his patch is now
no longer relevant since the section he was changing has now been
completely removed from the 3.2.1 version.)
Tested by: Alan W. Irwin <airwin@users.sourceforge.net> on Linux
by running the cmake command and making sure that wxwidgets was found.
M cmake/modules/FindwxWidgets.cmake
commit e23628eaa79ad88bd387eab6a4f3d46b8ebebbc4
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Tue Jun 2 19:45:37 2015 -0700
Build system: Work around linking issue for traditional static build
For the following conditions:
1. static build of PLplot libraries;
2. C++ code in libplplot (i.e., if any of the qt, psttf, or wxwidgets
devices are enabled); and
3. traditional build of PLplot examples that are written in languages
other than C or C++
it turns out that information on the required special C++ libraries
(e.g., libstdc++.so for the g++ compiler) is missing from the explicit
PLplot linker information that has been collected for our configured
*.pc files unless special measures are taken.
This solution for the issue is elegant because it simply relies on
CMake's knowledge of the C++ special libraries (held in
CMAKE_CXX_IMPLICIT_LINK_LIBRARIES and
CMAKE_CXX_IMPLICIT_LINK_DIRECTORIES) to configure our pkg-config files
correctly. We simply transform those CMake variables into an add-on
for the Libs.private line in plplot.pc. Thus, this additional
information has no effect unless the --static option is used when
pkg-config is called, in which case all other PLplot *.pc files access
the plplot.pc information via their "Requires.private: plplot" lines
(or "Requires: plplot" line in the case of plplot-ocaml.pc). The net
result of this is the static traditional build of the PLplot examples
should always include the special C++ libraries that CMake uses for
its own builds for all languages.
N.B. this commit drops all prior workarounds such as using the C++
compiler for the traditional build of the examples written in C (i.e.,
those in examples/c and examples/tk) and dropping all examples for the
static traditional build that are written in anything other than C or
C++. So the results for the traditional build and test in the static
case are much more substantial (include f95, ada, adathick and D)
than prior to this commit when those workarounds were in effect.
Tested by: Alan W. Irwin <airwin@users.sourceforge.net> on Linux
by running
scripts/comprehensive_test.sh --do_test_interactive no --do_ctest no \
--do_test_build_tree no --do_test_install_tree no \
--cmake_added_options "-DDEFAULT_NO_DEVICES=ON -DPLD_ps=ON \
-DPLD_psttf=ON"
This test previously generated errors before the last set of
workarounds for this issue were put into place, but now it
produces perfect results.
N.B. additional testing requested on the various Windows platforms
and Mac OS X.
M cmake/modules/pkg-config.cmake
M cmake/modules/plplot.cmake
M examples/CMakeLists.txt
M examples/c/CMakeLists.txt
M examples/tk/CMakeLists.txt
M src/CMakeLists.txt
commit 0e677d3831696cbcdd812490108e9434ea3b3fd7
Author: Jim Dishaw <jim@dishaw.org>
Date: Mon Jun 1 17:27:44 2015 -0400
Fixed uninitialized variable error in read_entry()
- Compiler complained that pdf_rc may be unintialized
M src/plmetafile.c
commit b2f689b4c5693008d232ec9c17f70d2885d2863a
Author: Jim Dishaw <jim@dishaw.org>
Date: Mon Jun 1 17:22:01 2015 -0400
Fixed uninitialized value warning in plreadmetafile()
- The compiler generated warning that plm is unintialized.
M src/plmetafile.c
commit b1217089d8ca92c29cfd090feaf979f8f3623556
Author: jdishaw <jim@dishaw.org>
Date: Mon Jun 1 14:24:17 2015 -0400
Ignore .DS_store and ._.DS_Store files
.DS_Store and ._.DS_Store files are used by MacOSX to store file
attributes. Ignore these files for git and also when our CMake-based
build system generates a source distribution tarball.
M .gitignore
M CMakeLists.txt
commit 92bff790efbc944603746d34b0c95d863e8bce67
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Sun May 31 21:56:54 2015 +0100
Added extra checks to rdbuf_bop to ensure we don't get an unassigned
command and call plabort if we have either no command or not a state
command.
M src/plbuf.c
commit 682273cc979c15c009e802a2a78124cf1b3ef05c
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Sun May 31 21:56:46 2015 +0100
Styled previous wxWidget binding commit.
M bindings/wxwidgets/wxPLplotwindow.h
commit f5b60f3696af773853d5bd0a0a245c1aabf2185a
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Sun May 31 17:59:23 2015 +0100
delete m_memoryDc and m_gcDc in the destructor of wxPLplotwindow to fix
the memory leak reported by Laurent Berger on 21 Apr 2015
M bindings/wxwidgets/wxPLplotwindow.h
commit e985bc321f12ac9a7f0edc667c13d31c29cad8fd
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Sun May 31 12:36:26 2015 +0100
Fixed wxWidgets driver issue whaere only the first subplot on a page was
being rendered. This turned out to be a bug in wxWidgets with wxGCDC.
This has now been reported - see http://trac.wxwidgets.org/ticket/17013
for details. I have put some workarounds in place and things should now
function as expected.
M drivers/wxwidgets_dev.cpp
commit 87a9504019bf06ab5792eb5eaf140d79ece7baf8
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sat May 30 22:56:14 2015 -0700
Build system: Fix qhull-related bugs
Changed FindQHULL.cmake to create a more robust
environment for finding the qhull header qhull_a.h (including looking for
that header in both the qhull and libqhull subdirectories) and the
qhull library.
Changed csiro.cmake to give more informative success or failure messages
for the qhull case.
To be consistent with the above header search must also change our
==>
Tested by: Alan W. Irwin <airwin@users.sourceforge.net> on Linux
by running
make -j4 x21c ; make xwin
and
examples/c/x21c -dev xwin
to assure all is well with our use of qhull.
M cmake/modules/FindQHULL.cmake
M cmake/modules/csiro.cmake
M lib/nn/delaunay.c
M src/plgridd.c
commit 7d1dfa81214800e2be19cd17f7ec895d2810ea17
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sat May 30 14:11:31 2015 -0700
Build system: tweak WARNING message when hdf5.h is not found
M cmake/modules/octave.cmake
commit a407d2476605d597c6bce52fc17554ae98175488
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Sat May 30 12:03:31 2015 +0100
Changed build tree checking in plcore after I messed up a merge conflict for the previous commit
M src/plcore.c
commit f38179a7b5289c95cd2c07e7d281ac0c29b57d97
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Sat May 30 11:37:52 2015 +0100
Styled previous commits
M drivers/wxwidgets.cpp
M drivers/wxwidgets_comms.h
M drivers/wxwidgets_dev.cpp
M src/plcore.c
M utils/wxplframe.cpp
commit 2e594f68c9708df0c45d2816945a2c1bc2801762
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Sat May 30 11:06:24 2015 +0100
Sped up the checking of text size in wxPLViewer by polling for the
result more often
M drivers/wxwidgets_dev.cpp
commit 0b0dc58326b41529f0b742afd965164d629f0255
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Sat May 30 10:23:06 2015 +0100
Fixed issue reported in previous commit. The cause was that for a
wxGCDC setting the clip region to a zero sized rectangle results in
clipping everything. The wxWidgets documentation states this should be
equivalent to removing the clipping region altogether. This is a problem
because wxDC::GetClippingBox returns a zero size rectangle if there is
no clipping, therefore if we have no clipping then checking the clipping
box, setting it, then restoring it to its original value results in
clipping everything if the original state was clipping nothing.
This is clearly a wxWidgets bug. I have worked around it by checking for
a zero sized box and performing the correct action.
M drivers/wxwidgets_dev.cpp
commit b54d86fe6e069c8aa656062c5a869bbf341c765f
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Sat May 30 10:09:40 2015 +0100
Added a clipper class to wxwidgets_dev.cpp. This differs from
wxDCClipper in that it can deal with (i.e. does nothing for) a null dc
and upon destruction it restores the clipping region to the previous
state rather than removing all clipping regions.
Removed all the if( m_dc ) checks from calls to the DC state changing
objects in wxPLDevice. See previous commit for details of why this was a
problem and modifications to these classes to allow null pointers.
Now we have text back for wxPLplotDemo, but still some issues in
wxPLViewer - for x00 we have no data in the plot and only a single axis
label of 0.0
M drivers/wxwidgets_dev.cpp
commit 22f1d0ca835555ba7977a6d10847acc58ee8a9c7
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Sat May 30 09:21:33 2015 +0100
Added checks for a null dc to all the "wxDC state changer" classes in
wxwidgets_dev.cpp. They need to be able to accept null becuase when
using wxPLViewer we want to reuse the same code for chunking text, but
we cannot put the state changers in an if ( m_dc ) statement otherwise
they go immediately out of scope and undo the changes that were made.
M drivers/wxwidgets_dev.cpp
commit 3ee3eb4b409cf5b8a073520135b6187314358817
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Fri May 29 15:25:29 2015 +0100
Added support for getting string length in wxWidgets, including for
wxPLViewer. For the case where a dc is passed to wxWidgets most of the
code was there anyway. For wxPLViewer this involved a new command passed
to the viewer and some extra data in the header to get the text size
from the viewer.
Somehow this commit has broken text rendering completely, but I'm not
sure how. I am making this commit so I can easily diff it to see what
has caused the problem.
M drivers/wxwidgets.cpp
M drivers/wxwidgets_comms.cpp
M drivers/wxwidgets_comms.h
M drivers/wxwidgets_dev.cpp
M utils/wxplframe.cpp
commit 818b93adce89a40901272bc766d046cf92893538
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Tue May 26 13:11:06 2015 -0700
Build system: exclude testing the wxwidgets device from the test_interactive target
The issue is the -np option does not yet work for the wxwidgets
device. Therefore, including this device in the list of tests run by
the test_interactive target significantly increases the interactive
load (clicking on the mouse to display all the pages and eventually
exit the wxPLViewer GUI for each of the examples that are tested) when
running that target, and thus makes running that target much too heavy
a burden for interactive testers.
M examples/CMakeLists.txt
M plplot_test/CMakeLists.txt
commit 3644bc341c2257ba90396302077a1220226173e9
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sat May 23 09:11:46 2015 -0700
Build system:libplplot: Improved fix for plInBuildTree on Windows
The previous IGNORECASE fix for plInBuildTree in the Windows case had
a CMake-level issue that IGNORECASE was not being set for the static
build case on Windows. This bug fix solves that CMake-level issue and
therefore allows reversion of a recently implemented C-level fix for
the non-IGNORECASE code path that in any case was not general enough
(didn't handle case issues on Windows like the IGNORECASE C-level code
path does).
Tested by Alan W. Irwin <airwin@users.sourceforge.net> on Linux using
scripts/comprehensive_test.sh. N.B. Additional testing on both
MSVC and non-MSVC (Cygwin, MinGW/MSYS) windows platforms is needed.
M src/CMakeLists.txt
M src/plcore.c
commit 7c565e369450d311bc1b3cc5f2a39e271549e732
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sat May 23 09:07:31 2015 -0700
Style previous commit
M drivers/wxwidgets_dev.cpp
commit 38ea06417dbafa0a5183c3e1f7a1379aa32be6eb
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Fri May 22 17:42:38 2015 -0700
Build system: replace use of the transform_version function with VERSION_LESS
I wrote that transform_version function years ago (before VERSION_LESS
was available for version comparisons of arbitrary depth) to transform
version triplets into a number that could then be evaluated with LESS
logic. It has proven reasonably reliable up to now when Orion
discovered it choked on a version string of 2.007 for PDL on Fedora
because of the unusual number of components (only two). I then did
tests of VERSION_LESS, and it is vastly superior and handles, e.g.,
comparisons with 2.007 with ease. Therefore, I have now dropped
defining and using transform_version, and have converted all the
surrounding logic to simple VERSION_LESS comparisions which should fix
the problem that Orion has encountered on Fedora.
Tested by: Alan W. Irwin <airwin@users.sourceforge.net> on Linux
by running CMake and making sure there were no extra warnings
concerning version checks that did not pass.
M cmake/modules/cairo.cmake
M cmake/modules/lua.cmake
M cmake/modules/octave.cmake
M cmake/modules/pdl.cmake
M cmake/modules/plplot.cmake
M cmake/modules/plplot_functions.cmake
M cmake/modules/python.cmake
commit 7e6b860556402a9648f30e1f67a7a4608113e580
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Fri May 22 11:23:44 2015 +0100
Added missing counter increment when waiting for wxPLViewer to signal it
is ready
M drivers/wxwidgets_dev.cpp
commit f5dc38920e3118e5abddc07686af04347b765ddb
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Fri May 22 11:17:47 2015 +0100
Fixed check for being in the build directory on windows. Windows uses \
for its path separator not / so if building an example the check failed
which in my case meant that the font files couldn't be opened and plplot
exited.
M src/plcore.c
commit b077fd9c09e7ba2df7691e5c7bb96d8b5db0cb50
Author: Phil Rosenberg <p.d.rosenberg@gmail.com>
Date: Fri May 22 10:44:18 2015 +0100
Set the wxWidget OriginChanger to use wxCoord, rather than long. Use of
long for coordinates is deprecated as of wxWidgets 3.0
M drivers/wxwidgets_dev.cpp
commit c4ba8701c528f7ae4ee1b3588a5a3c87cfe103c9
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Thu May 21 12:01:13 2015 -0700
Build system: Use cmake_minimum_required in Ada language support
This fix should quiet (I hope!) the last Cygwin warning concerning
"CMake no longer defines WIN32 on Cygwin!". In this case the warning
was actually correct and not spurious; project was being called in the
Ada language support files without any preceding call to
cmake_minimum_required before this fix.
M cmake/modules/language_support/cmake/CMakeTestAdaCompiler.cmake
M cmake/test_ada/cmake_experimental/Modules/CMakeTestAdaCompiler.cmake
M cmake/test_ada/cmake_working/Modules/CMakeTestAdaCompiler.cmake
commit a09dc18c886cb0e9a2b2ada4c6e836a181c2a887
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed May 20 21:16:49 2015 -0700
Build system: Work around linking issue for traditional static build
For the following conditions:
1. static build of PLplot libraries;
2. C++ code in libplplot (i.e., if any of the qt, psttf, or wxwidgets
devices are enabled); and
3. traditional build of PLplot examples that are written in languages
other than C or C++
it turns out that information on the required special C++ library
(e.g., libstdc++.so for the g++ compiler) is missing from the explicit
PLplot linker information (because cannot work around this problem by
linking with g++, for e.g., the f95 examples). Therefore, the
traditional build sometimes fails for this special combination of
circumstances. (It also sometimes succeeds by accident on some
platforms if some shared system library such as the Qt libraries is
linked which drags in a link to the correct version of the special C++
library, IF, there is a match in g++ special library version between
the system build of Qt and the PLplot build). But, for example, this
does not happen on Linux if psttf is the only C++ device included, and
that external library happens to be only available in static form. So
my preliminary test of that case failed demonstrating the problematic
nature of the linking of the traditional build for this special
combination of circumstances.
Note, CMake is smart enough to figure out this difficult case
(presumably by figuring out the names and locations of the special
libraries associated with each kind of compiler). And the traditional
build system can fudge the correct special C++ library by using the
C++ compiler for all the C example builds. But figuring out the name
and location of the C++ library for all C++ compilers that CMake
supports is far beyond what can be expected from our traditional build
system. Therefore, exclude the non-C or C++ compiled language examples
(i.e., ada, d, f95, java, and ocaml) of the traditional comprehensive
test for this combination of circumstances.
M examples/CMakeLists.txt
M examples/Makefile.examples.in
M examples/ada/Makefile.examples.in
M examples/d/Makefile.examples.in
M examples/f95/Makefile.examples.in
M examples/java/Makefile.examples.in
M examples/ocaml/Makefile.examples.in
commit 0c32ea9a2c245f6c6d5a41f399f9c3106570478d
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed May 20 17:44:33 2015 -0700
Build system: set target property POSITION_INDEPENDENT_CODE ON for all PLplot libraries
This change should allow shared libraries from foreign software
packages (or user's code) to link properly to static PLplot libraries.
The cost of this gain is there is a small amount of inefficiency
(e.g., the more complicated addressing code that apparently occurs on
Linux for all code compiled with the -fPIC flag) introduced for the
static PLplot libraries.
Tested by: Alan W. Irwin <airwin@users.sourceforge.net> on Linux using
scripts/comprehensive_test.sh --do_ctest no --do_test_interactive no --do_test_build_tree no
M cmake/modules/plplot_functions.cmake
commit 6ef04daa5853947698a5597a342f18392ae2d3b3
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed May 20 16:27:17 2015 -0700
Build system: Simplify the logic by implementing and using set_library_properties function
To simplify the build system logic and to improve the uniformity of
the way library target properties are set, implement and use the
set_library_properties function which sets all the usual target properties
for a given PLplot library.
Tested by: Alan W. Irwin <airwin@users.sourceforge.net> on Linux
using
scripts/comprehensive_test.sh --do_ctest no --do_test_interactive no --do_test_build_tree no
M bindings/ada/CMakeLists.txt
M bindings/c++/CMakeLists.txt
M bindings/d/CMakeLists.txt
M bindings/f95/CMakeLists.txt
M bindings/qt_gui/CMakeLists.txt
M bindings/tcl/CMakeLists.txt
M bindings/wxwidgets/CMakeLists.txt
M cmake/modules/instdirs.cmake
M cmake/modules/plplot_functions.cmake
M cmake/modules/plplot_version.cmake
M lib/csa/CMakeLists.txt
M lib/nistcd/CMakeLists.txt
M lib/nn/CMakeLists.txt
M lib/qsastime/CMakeLists.txt
M src/CMakeLists.txt
commit 1a20a391f83d2076a220dc23a3286f1ef53804f3
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed May 20 11:50:15 2015 -0700
Build system: fix logic problems for the case when devices are limited
Without this fix, "-DDEFAULT_NO_DEVICES=ON -DPLD_ps=ON -DPLD_psttf=ON"
did not work correctly.
Tested by: Alan W. Irwin <airwin@users.sourceforge.net> on Linux using
scripts/comprehensive_test.sh \
--do_test_interactive no \
--do_ctest no \
--do_test_build_tree no \
--do_test_install_tree no \
--cmake_added_options "-DDEFAULT_NO_DEVICES=ON -DPLD_ps=ON -DPLD_psttf=ON"
M cmake/modules/qt.cmake
M examples/c/CMakeLists.txt
commit 3062c3ba91b02a323c56e8987fe9f1360ba30abf
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Mon May 18 16:50:12 2015 -0700
Build system: change pkg-config *.pc file default install location
My previous choice of ${CMAKE_INSTALL_DATADIR}/pkgconfig for the
default value of CMAKE_INSTALL_PKG_CONFIG_DIR is an architecture-
independent location, but as pointed out by Orion the contents of
these *.pc files very much depend on architecture. Therefore, I have
changed the default install location to
+${CMAKE_INSTALL_LIBDIR}/pkgconfig which has the added advantage that
most disto packagers will only need to override CMAKE_INSTALL_LIBDIR
alone to get the correct CMAKE_INSTALL_PKG_CONFIG_DIR result without
having to override CMAKE_INSTALL_PKG_CONFIG_DIR as well.
M cmake/modules/instdirs.cmake
commit 1424994fcf6db505883e87ac9029929c83913b61
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Mon May 18 11:37:20 2015 -0700
Core library: Add needed level check for calls to plP_state
Put a
if ( plsc->level > 0 )
level check on all calls to
plP_state( PLSTATE_FILL );
in src/plsdef.c. Without these level checks, then the combination
of
plschr( 0.0, 1.0);
plinit();
in that order leads to an invalid read as reported by valgrind and
a segfault. See also <https://sourceforge.net/p/plplot/bugs/162/>.
Note the wxwidgets/plbuf changes associated with the last release
cycle moved a call to plP_state to the spat routine (but without the
level check that it had in its previous location). So that was a
regression against 5.10.0. Furthermore, the wxwidgets/plbuf changes
introduced two additional calls to plP_state within this source file
without the level check. So this fix for those three cases solves a
regression and two new bugs introduced by the wxwidgets/plbuf changes.
M src/plsdef.c
commit e8520ea7068dde441c63aa6a1160ac290ab909e3
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Mon May 18 11:11:54 2015 -0700
Fix styling issue that caused a bad result for scripts/style_source.sh
The bad result was in include/pldll.h.in where for an #if preprocessor
directive uncrustify removed a misaligned continuation "\" character
rather than aligning it on the end of a second line of a whole bunch
of continued lines. The solution was to align the "\" by hand, and
after that change uncrustify no longer removed the continuation
character.
N.B. This is the first time in years that uncrustify has introduced a
code error, but the episode is useful because it should strongly
motivate us to always carefully check what uncrustify does (typically
with "git diff --ignore-all-space") before committing the result.
M include/pldll.h.in
commit 2b4e39742da82606e93b8bcc34ff22e58a6bb84e
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Mon May 18 00:03:24 2015 -0700
Build system: Adopt new user-configurable install location CMAKE_INSTALL_PKG_CONFIG_DIR for PLplot-generated pkg-config *.pc files
Tested by Alan W. Irwin <airwin@users.sourceforge.net> on Linux using
scripts/comprehensive_test.sh --do_test_interactive no --do_ctest no --do_test_build_tree no --do_test_install_tree no
M cmake/modules/instdirs.cmake
M cmake/modules/pkg-config.cmake
commit d0464adff50ac79421fad50bd3263531da55bd1d
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sun May 17 21:34:29 2015 -0700
Build system: Drop problematic specific Cygwin logic for the traditional build of the fortran examples.
This problematic logic was introduced in 2007 for f77 and inherited
for f95 later. It specifies for the traditional build a versioned
library name for both libplplotf95c and libplplotf95 installed in a
non-standard location. This problematic logic may have worked back in
2007, but it definitely does not work now leading to traditional build
errors on Cygwin with a message stating "cannot find
-lplplotf95-12.0.0".
So as an experiment I have completely removed this logic and instead
treat the traditional build of fortran examples in the same standard
way as the C, C++, etc., examples.
N.B. needs testing on Cygwin to see whether removing this problematic
logic for Cygwin is the right approach.
M bindings/f95/CMakeLists.txt
commit 417eeb1e129d0feb99f4f4a0ecbc45b824b60cfd
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sun May 17 21:09:19 2015 -0700
Traditional build system for installed examples: Fix determination of extXdrawable_true and extcairo_true
These two variables used to configure examples/c/Makefile.examples
are now being set using exactly the same logic as the CMake build
system. This should fix a cross-platform issue (which Arjen happened to
discover on his Cygwin platform) where these variables were being
set inconsistently with the availability of required pkg-config
packages for gtk+-x11-2.0 and/or cairo leading to build errors
for the traditional build system.
M cmake/modules/cairo.cmake
M examples/c/CMakeLists.txt
commit 2dbd4f263b084633b81c0beb7210b716f949f264
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sun May 17 21:05:49 2015 -0700
Build system: Fix linking issues with traditional build of installed examples (update)
Fix an issue where pkg-config style variables to configure
Makefile.examples would not be set unless BUILD_TEST was true
for the core build.
M examples/c++/CMakeLists.txt
M examples/tk/CMakeLists.txt
commit 782b09c16ab13672669b0b8e4790820ce1aa5009
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sun May 17 17:04:44 2015 -0700
Build system cleanup; remove all references to gnome device
The gnome device (since superseded by the gcw/gnome2/plplotcanvas
effort which has also been retired a long time and recently cleaned
out) has been retired since 2006 with no users requests to restore it
so run
find . -type f |grep -v .git |xargs grep -li gnome
to find all files which contain this term, and edit those files to
clean out everything associated with the gnome device.
Tested by: Alan W. Irwin <airwin@users.sourceforge.net> on Linux using
scripts/comprehensive_test.sh --do_test_interactive no
(The interactive portion of the tests were dropped as a matter of
convenience.)
M bindings/tk/about.tcl
M cmake/modules/drivers-init.cmake
M cmake/modules/drivers.cmake
M include/drivers.h
M include/plDevs.h.in
M include/plcore.h
M plplot_config.h.in
M plplot_test/plplot-test.sh.in
commit 94932d47cd13c146c35207e3f6d42dcd419bc063
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sun May 17 15:19:01 2015 -0700
Build system cleanup; remove all references to gcw, gnome2, and plplotcanvas
The gcw device and associated gnome2 bindings and plplotcanvas examples
have been retired since 2010 with no users requests to restore them
so run
find . -type f |grep -v .git |xargs grep -E -li '(gcw|gnome2|plplotcanvas)'
to find all files which contain those terms, and either edit or remove
the files as appropriate to clean out everything associated with those
terms. Note this is quite an intrusive change, but see test below.
Tested by: Alan W. Irwin <airwin@users.sourceforge.net> on Linux
using
scripts/comprehensive_test.sh --do_test_interactive no
(The interactive portion of the tests were dropped as a matter
of convenience.)
M abi-compliance-checker.xml.template.in
M bindings/CMakeLists.txt
D bindings/gnome2/CMakeLists.txt
D bindings/gnome2/README
D bindings/gnome2/lib/CMakeLists.txt
D bindings/gnome2/lib/gcw-lib.c
D bindings/gnome2/lib/plplotcanvas.c
D bindings/gnome2/python/CMakeLists.txt
D bindings/gnome2/python/README
D bindings/gnome2/python/cplplotcanvas.override
D bindings/gnome2/python/cplplotcanvasmodule.c
D bindings/gnome2/python/gcw.override
D bindings/gnome2/python/gcwmodule.c
D bindings/gnome2/python/plplotcanvas.py
M cmake/modules/drivers-init.cmake
M cmake/modules/drivers.cmake
D cmake/modules/gcw.cmake
M cmake/modules/plplot_version.cmake
M drivers/deprecated_wxwidgets.h
D drivers/gcw.c
D drivers/gcw.driver_info.in
D drivers/plplotcanvas-hacktext.c
M examples/CMakeLists.txt
M examples/c/CMakeLists.txt
M examples/c/Makefile.examples.in
D examples/c/README.plplotcanvas
D examples/c/plplotcanvas_animation.c
D examples/c/plplotcanvas_demo.c
M examples/plplot_configure.cmake_installed_examples.in
M examples/python/CMakeLists.txt
D examples/python/README.plplotcanvas
D examples/python/plplotcanvas_animation.py
D examples/python/plplotcanvas_demo.py
M include/CMakeLists.txt
M include/drivers.h
D include/gcw.h
M include/plDevs.h.in
M include/plcore.h
M include/pldll.h.in
D include/plplotcanvas-hacktext.h
D include/plplotcanvas.h
M plplot_test/plplot-test-interactive.sh.in
M plplot_test/plplot-test.sh.in
M scripts/style_source.sh
M src/plbuf.c
M www/credits.php
commit 6a5bad583d46dae088f1bfd8372eb30b9dcbe762
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sun May 17 00:43:37 2015 -0700
Build system: Fix linking issues with traditional build of installed examples
The general issues fixed in this commit are the following:
1. Distinctions between PUBLIC (for shared libraries) and PRIVATE for
(static libraries) linking were not being done in the configured *.pc
files leading to overlinking issues for the shared libraries case.
Now, most libraries (unless they are always supposed to be PUBLIC) are
assigned to PUBLIC or PRIVATE depending on whether NON_TRANSITIVE is
OFF or ON.
2. The linking configuration of the *.pc files was additionally
incomplete for some of our libraries leading to underlinking.
3. The link steps in the Makefiles for the traditional build were
being configured to an incomplete result leading to underlinking.
I dealt with these issues in general by transforming the linking
solutions developed in the CMake case (since those are better debugged)
to the pkg-config and configured Makefile case.
Along the way, I discovered a fixed the following issues in our
CMake build system:
A. libplplotada must always link to libplplot as PUBLIC.
B. pkg_config_link_flags function must allow empty items in the lists of
libraries it translates into pkg-config form. (This was handled
by using quotes appropriately.)
C. Some libraries need to be treated as PUBLIC regardless (such as
case A above). To handle this issue for the pkg-config case,
I added an optional 7th argument to pkg_config_file containing
a list of libraries that should be treated as always PUBLIC.
D. --static option now used for pkg-config results in the traditional
build when BUILD_SHARED_LIBS=OFF. (Previously this only happened when
FORCE_EXTERNAL_STATIC was ON which was a mistake since that
experimental option is used for another purpose.)
Tested by Alan W. Irwin <airwin@users.sourceforge.net> on Linux
using default (i.e., everything tested) scripts/comprehensive_test.sh.
N.B. this only tested the Qt4 case on Debian stable so additional
comprehensive testing is requested for Qt5 and/or additional platforms.
M bindings/ada/CMakeLists.txt
M bindings/c++/CMakeLists.txt
M bindings/d/CMakeLists.txt
M bindings/f95/CMakeLists.txt
M bindings/qt_gui/CMakeLists.txt
M bindings/tcl/CMakeLists.txt
M bindings/wxwidgets/CMakeLists.txt
M cmake/modules/pkg-config.cmake
M examples/CMakeLists.txt
M examples/ada/Makefile.examples.in
M examples/c++/CMakeLists.txt
M examples/c++/Makefile.examples.in
M examples/c/Makefile.examples.in
M examples/d/Makefile.examples.in
M examples/f95/CMakeLists.txt
M examples/f95/Makefile.examples.in
M examples/tk/CMakeLists.txt
M examples/tk/Makefile.examples.in
M pkgcfg/plplot-template.pc.in
commit efc9dfacc9e33c00aa01d150a50d03cfd80dc099
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sun May 17 00:00:00 2015 -0700
Core C library: fix floating point issue in notcrossed function (second try)
The last commit concerning this topic had some problems with
uninitialized variables. The results of calculations with those
uninitialized variables were ignored (because status !=0) so there
were no practical consequences. Presumably, however, those ignored
calculations would trigger additional floating-point exceptions by
chance. This commit cleans up all that logic so that uninitialized
variables are never used for calculations in the notcrossed function.
M src/plfill.c
commit e2995e9f8df11ddb4abddd3b4e818b850b091b40
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Fri May 15 14:39:17 2015 -0700
Style previous commit
M src/plcont.c
commit cb528a2bde8fe40ffeee98e0e115a7a9a99b2f86
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Fri May 15 12:56:03 2015 -0700
Build system: Update implementation of different cmake minimum versions on different platforms
In this new simplified implementation drops the idea of using a
function (since Brad King has just generated documentation stating
that calling cmake_minimum_version inside a function is problematic)
and instead simply use cmake_policy(VERSION 3.0.2) to enforce uniform
policy for different minimum CMake versions on different platforms.
Tested by: Alan W. Irwin <airwin@users.sourceforge.net> on Linux by
configuring PLplot with various versions of CMake. These tests
generated no policy warnings or any other unexpected error or warning
messages.
N.B. this change still needs testing on Cygwin, but my hope is this
change finally squelches the last of the spurious warning messages on
that platform.
M CMakeLists.txt
M examples/CMakeLists.txt
commit d41df4f8a1c0cd7db5f6b227347910e8b9363e8a
Author: Arjen Markus <arjen.markus@deltares.nl>
Date: Fri May 15 13:21:15 2015 +0200
Add interfaces for the date/time handling routines. One issue remaining: the boolean argument in plconfigtime.
M bindings/f95/sfstubsf95.f90
commit 793c0b6cd331ce58379829eb9958363e8abeb799
Author: Hazen Babcock <hbabcock@fas.harvard.edu>
Date: Thu May 14 11:46:02 2015 -0400
Fix sign problem.
M src/plcont.c
commit 7ae509dd7692b62b9f8ab0592c72cdd7071d11d0
Author: Hazen Babcock <hbabcock@fas.harvard.edu>
Date: Thu May 14 11:37:57 2015 -0400
Change sign calculation logic in pldrawcn() to avoid floating point under / over flow during calculation of the iedge array.
M src/plcont.c
commit a0e51cde2b1c4234f9e043a09fe0f226d69557e0
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed May 13 16:58:56 2015 -0700
Fortran example 29: fix argument specification bug
The change was 0. ==> 0._plflt in two different plctime calls and this
cleared up the case where the seconds variable was being replaced by a
random floating point number in the C version of the plctime call.
This change also apparently fixed the floating-point exceptions for
Fortran example 29 although I leave it to Hazen (who drew attention
to the plctime issue) to explain why.
M examples/f95/x29f.f90
commit b916d4bcbc051712e2eb1f57c195cabdb7140f4a
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed May 13 14:21:38 2015 -0700
Core C library: fix floating point issue in notcrossed function
The issue was that sometimes the calculated intersection was far
outside the line segments so the conversion of that floating-point
information to PLINT was out of range. The returned intersection is
only used when status = 0 (i.e., there was a definite crossing within
both line segments) so the fix was only to do that conversion when
status = 0.
This change fixed the floating-point issues revealed by using
the gfortran option -ffpe-trap=invalid,zero,overflow,underflow'
for fortran examples 25 and 30, but floating-point issues still
remain for examples 21 and 33.
M src/plfill.c
commit c67fd667347b09bf5d3be6c0bc73de3762e59ab1
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Tue May 12 01:47:36 2015 -0700
Build system: replace deprecated use of LINK_INTERFACE_LIBRARIES with recommended INTERFACE_LINK_LIBRARIES
Policies CMP0022 and CMP0023 have been set to NEW, and all use of
LINK_INTERFACE_LIBRARIES has been replaced by using either the
PRIVATE (NON_TRANSITIVE=ON, INTERFACE_LINK_LIBRARIES set to empty) or
PUBLIC (NON_TRANSITIVE=OFF, INTERFACE_LINK_LIBRARIES set to listed
libraries) target_link_libraries keywords.
Tested by: Alan W. Irwin <airwin@users.sourceforge.net> on Linux
by running
scripts/comprehensive_test.sh
Furthermore, for the shared library case, I have selectively used the
"ldd -u -r" command on various executables, drivers, and libraries to
show that overlinking is almost entirely removed by setting
NON_TRANSITIVE=ON (which happens by default on Linux).
M CMakeLists.txt
M bindings/ada/CMakeLists.txt
M bindings/c++/CMakeLists.txt
M bindings/d/CMakeLists.txt
M bindings/f95/CMakeLists.txt
M bindings/qt_gui/CMakeLists.txt
M bindings/tcl/CMakeLists.txt
M bindings/wxwidgets/CMakeLists.txt
M examples/CMakeLists.txt
M lib/csa/CMakeLists.txt
M lib/nn/CMakeLists.txt
M lib/qsastime/CMakeLists.txt
M src/CMakeLists.txt
commit 80bacaeb3fc9ca5a19b6c4045a89f7de7a030aff
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Mon May 11 16:04:08 2015 -0700
Build system: Finish change to call cmake_minimum_required command before project command
This part of the change affects the installed examples build and the
soft-landing logic for compiler detection and should finish this
effort for our PLplot builds. This reording should quiet the spurious
Cygwin warnings about "CMake no longer defines WIN32 on Cygwin!" and
also solve some other potential issues when cmake_minimum_required
and project commands are called in the incorrect order (as occurred
before this fix).
Tested by: Alan W. Irwin <airwin@users.sourceforge.net> on Linux
using
scripts/comprehensive_test.sh \
--do_test_interactive no \
--do_nondynamic no \
--do_static no
Further testing (especially on Cygwin to confirm the warnings are
gone) is requested.
M cmake/modules/language_support.cmake
M examples/CMakeLists.txt
commit 0f1adc221c27e0b8f342398f0862625e5b9a0840
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Mon May 11 11:40:23 2015 -0700
comprehensive_test.sh: collect CMake cache files in tarball.
M scripts/comprehensive_test.sh
commit 29ae9bf98aa0f4ce475241bd1cb7349b835ce821
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Mon May 11 10:43:25 2015 -0700
Traditional build of installed examples: Fix test_tcl_psc target
The (non-Windows) issue was the results variable was being set to "." rather than
the full path corresponding to OUTPUT_DIR=. so the results were being
put into the incorrect (tcl) directory.
Tested: by Alan W. Irwin <airwin@users.sourcforge.net> on Linux
using
scripts/comprehensive_test.sh --cmake_added_options \
"-DDEFAULT_NO_DEVICES=ON -DPLD_tk=ON -DPLD_xwin=ON -DPLD_ntk=ON -DPLD_tkwin=ON -DPLD_ps=ON -DDEFAULT_NO_BINDINGS=ON -DENABLE_tcl=ON -DENABLE_itcl=ON -DENABLE_tk=ON -DENABLE_itk=ON" \
--do_test_interactive no
M plplot_test/test_tcl.sh.in
commit 7f9ba0937f237c25ce00dcfda17ce117e96dbce4
Author: Aaron Hexamer <hexamer@comcast.net>
Date: Sat May 9 08:21:34 2015 -0500
Build system: fix qt configuration
Deal properly with the case where memqt is the only Qt device.
Consistently use PLD_extqt rather than ENABLE_qt to identify when
the pyqt4 bindings should be implemented and the qt_example
application built.
Tested by: Alan W. Irwin <airwin@users.sourceforge.net> on Linux
using the following combinations of cmake options and tests:
1. extqt only (-DDEFAULT_NO_QT_DEVICES=ON -DPLD_extqt=ON)
Built test_all_qt (which runs the C++ and pyqt4 forms of qt_example)
in both the build tree and installed examples tree.
2. memqt only (-DDEFAULT_NO_QT_DEVICES=ON -DPLD_memqt=ON)
2a. In the build tree, built the test_python_psc target (to populate all python
dependencies) and the mem, cairo, and qt targets (to populate
the mem, memcairo, and memqt devices). Then ran
examples/python/test_plsmem.py
in the build tree which generated correct mem, memcairo, and memqt results.
2b. For the CMake-based build of the installed examples, installed, then
ran
cmake $prefix/share/plplot5.11.0/examples
without options, then built the test_python_psc target (probably
not necessary), then ran
$prefix/share/plplot5.11.0/examples/python/test_plsmem.py
which generated correct mem, memcairo, and memqt results.
M bindings/qt_gui/plqt.cpp
M cmake/modules/drivers-init.cmake
M cmake/modules/qt.cmake
M examples/c++/CMakeLists.txt
M include/CMakeLists.txt
M include/qt.h
commit fc80e78681d36575b9930c2064f41115735c5542
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Fri May 8 23:22:09 2015 -0700
Build system: move to using CMP0026 NEW
This is a rather intrusive build-system change where use
of the deprecated LOCATION property for build targets is replaced by
using appropriate generator expressions.
One large complication of this change is that certain files now have
to be configured that contain generator expressions. I have designed
the configure_file_generate function to do that task which combines
appropriate calls to configure_file and file(GENERATE...) to configure
files containing both the normal configure_file tokens but also
generator expressions. This approach works fine for CMake 3.2, but
for versions less than that there is a file(GENERATE...) permissions
bug which has to be fixed with a GNU extension to chmod which is
apparently only available on Linux and Cygwin. But that dovetails
neatly with our desire to maintain a CMake minimum version of 3.0.2
for likely the next two years for both those platforms so it is only
for those platforms where a permission fix is convenient (using the
GNU extension to chmod) that a permissions fix is required.
Tested by Alan W. Irwin <airwin@users.sourceforge.net> on Linux
using
scripts/comprehensive_test.sh --cmake_added_options \
"-DDEFAULT_NO_DEVICES=ON -DPLD_tk=ON -DPLD_xwin=ON -DPLD_ntk=ON -DPLD_tkwin=ON -DPLD_ps=ON -DDEFAULT_NO_BINDINGS=ON -DENABLE_tcl=ON -DENABLE_itcl=ON -DENABLE_tk=ON -DENABLE_itk=ON"
The restrictions on the tests specified by --cmake_added_options are
to speed the test while not compromising the ability to comprehensively
test all Tcl and Tk targets which are the only ones where the
configure_file_generate function now has to be used.
M bindings/tcl/CMakeLists.txt
M bindings/tk-x-plat/CMakeLists.txt
M bindings/tk/CMakeLists.txt
M cmake/modules/plplot_functions.cmake
M cmake/modules/qt.cmake
M drivers/CMakeLists.txt
M examples/CMakeLists.txt
M examples/tcl/CMakeLists.txt
M examples/tk/CMakeLists.txt
M lib/nistcd/CMakeLists.txt
M plplot_test/CMakeLists.txt
M plplot_test/test_tcl.sh.in
commit 6c847e28cc9b5f2240d8ed174f723e8065cdb393
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Fri May 8 23:03:51 2015 -0700
Build system: Update minimum CMake version to 3.2.2 for most platforms
The platforms where the minimum version remains at 3.0.2 are Linux
and Cygwin, and that will likely continue for a couple of more years
until all modern Linux distros and Cygwin install a cmake version >= 3.2
The reason for requiring 3.2.2 for most platforms is there is a
permissions bug in file(GENERATE...) that is fixed in cmake 3.2, and a
straightforward workaround for that bug is only available for Linux
and Cygwin (e.g., not Mac OS X) where we are retaining the lower
minimum version in any case.
This commit also processes the plplot_test subdirectory before the
examples subdirectory so that the planned workaround_file_generate_bug
target which is designed to fix the permissions problem will work properly.
M CMakeLists.txt
commit 61494c103e2488ece708768365aecd0d5dec5b59
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Fri May 8 22:46:17 2015 -0700
Update scripts/comprehensive_test.sh to collect selected results in a tarball
That tarball (named comprehensive_test.tar.gz) is located in the
prefix directory and contains
1. Environment variable output.
2. Script output
3. All */output_tree files
4. A complete listing of all files.
That is, that tarball contains the complete information ordinarily
needed to report success or failure of the comprehensive_test.sh
script.
M scripts/comprehensive_test.sh
commit 0b43b1754669cf8388140d8b3d581f06236a2f77
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Fri May 8 13:13:58 2015 -0700
Title and comment tweaks for examples/tk/tk01.in and examples/tk/tk02.in
M examples/tk/tk01.in
M examples/tk/tk02.in
commit 3cb62f101f57b90a1642d687deca51509971f8b6
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Fri May 8 10:26:21 2015 -0700
Build system: Remove unused examples/tk/CMakeLists.txt_installed_examples_tk
D examples/tk/CMakeLists.txt_installed_examples_tk
commit d73609f18f024e4d293b34190cf138e5594578f7
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Thu May 7 17:22:39 2015 -0700
Release notes: Plot labels as date / time are now available for 3D plots
M README.release
commit b6a7ccd53860dbf69e44281591411effe46afb96
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Thu May 7 16:59:10 2015 -0700
DocBook documentation: pltimefmt
Replace the previous fairly minimal and sometimes incorrect
documentation of pltimefmt with complete documentation based on the
strfMJD implementation in lib/qsastime/qsastime.c and the original
documentation of that routine in lib/qsastime/qsastime.txt.
Tested by Alan W. Irwin <airwin@users.sourceforge.net> on Linux by
using the -DBUILD_DOC=ON cmake option, building the print target, and
viewing the resulting doc/docbook/src/plplot-5.11.0.pdf file in the
build tree.
M doc/docbook/src/api.xml
commit 433b8b2e59cb35ef031316197b6681a91eefe5c9
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Thu May 7 16:57:19 2015 -0700
Tweak commentary.
M src/plbox.c
commit 78e0d431ae4ce78d26e4a890c753b278a8011028
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Wed May 6 13:44:13 2015 -0700
libplplot: Add date/time formatting to plbox3
Tested by Alan W. Irwin <airwin@users.sourceforge.net> on
Linux using the following local change to the 11th standard
C example.
diff --git a/examples/c/x11c.c b/examples/c/x11c.c
index eca0fab..85d11a0 100644
--- a/examples/c/x11c.c
+++ b/examples/c/x11c.c
@@ -103,6 +103,7 @@ main( int argc, const char *argv[] )
z[i][j] = 3. * ( 1. - xx ) * ( 1. - xx ) * exp( -( xx * xx ) - ( yy + 1. ) * ( yy + 1. ) ) -
10. * ( xx / 5. - pow( xx, 3. ) - pow( yy, 5. ) ) * exp( -xx * xx - yy * yy ) -
1. / 3. * exp( -( xx + 1 ) * ( xx + 1 ) - ( yy * yy ) );
+ z[i][j] = 1.e4 * z[i][j];
if ( 0 ) // Jungfraujoch/Interlaken
{
@@ -112,12 +113,21 @@ main( int argc, const char *argv[] )
}
}
+ for ( i = 0; i < XPTS; i++ )
+ {
+ x[i] = 1.e4 * x[i];
+ }
+ for ( i = 0; i < YPTS; i++ )
+ {
+ y[i] = 1.e4 * y[i];
+ }
plMinMax2dGrid( (const PLFLT * const *) z, XPTS, YPTS, &zmax, &zmin );
step = ( zmax - zmin ) / ( nlevel + 1 );
for ( i = 0; i < nlevel; i++ )
clevel[i] = zmin + step + step * i;
cmap1_init();
+ pltimefmt( "%H:%M" );
for ( k = 0; k < 2; k++ )
{
for ( i = 0; i < 4; i++ )
@@ -126,10 +136,10 @@ main( int argc, const char *argv[] )
plcol0( 1 );
plvpor( 0.0, 1.0, 0.0, 0.9 );
plwind( -1.0, 1.0, -1.0, 1.5 );
- plw3d( 1.0, 1.0, 1.2, -3.0, 3.0, -3.0, 3.0, zmin, zmax, alt[k], az[k] );
- plbox3( "bnstu", "x axis", 0.0, 0,
- "bnstu", "y axis", 0.0, 0,
- "bcdmnstuv", "z axis", 0.0, 4 );
+ plw3d( 1.0, 1.0, 1.2, -3.0e4, 3.0e4, -3.0e4, 3.0e4, zmin, zmax, alt[k], az[k] );
+ plbox3( "bdnstu", "x axis", 0.0, 0,
+ "bdnstu", "y axis", 0.0, 0,
+ "bcdemnstuv", "z axis", 0.0, 4 );
plcol0( 2 );
This change simply scales x, y, and z by 1.e4 (so the ranges
correspond to several hours) and reformats all axis labelling using
date/time using the d suboption of xopt and yopt and the e suboption
of zopt. The resulting Hours:Minutes format looked good on all
3 axes.
M doc/docbook/src/api.xml
M src/plbox.c
commit 6ba58a92f515e3552eb8a9d6e1ddf967f25f151d
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Thu Apr 23 14:04:20 2015 -0700
test_diff.sh: bugfix: Replace psc by ${device} for stdout differences
Replaced some psc-specific logic for checking the stdout differences
with more general ${device} logic so that when the test device is
something other than psc, this script does not give spurious
missing stdout messages.
Tested by Alan W. Irwin <airwin@users.sourceforge.net> on Linux
by using the CMake option -DTEST_DEVICE:STRING=svg, running the all
target to fill in build system dependencies for ctest, then running
ctest -j4 --tests-regex '(compare|examples_(c|ocaml|ada))'
(that regex limits the scope of the test to comparisons of c++, ocaml, and
ada results with c results). The results were
[...]
26: Test command: /bin/bash "-c" "/home/software/plplot/HEAD/build_dir/plplot_test/test_diff.sh --device=svg"
26: Test timeout computed to be: 1500
26:
26: Comparison test using svg device
26:
26: c++
26: Missing examples :
26: Differing graphical output :
26: Missing stdout :
26: Differing stdout :
26: ada
26: Missing examples :
26: Differing graphical output : 08 19
26: Missing stdout :
26: Differing stdout :
26: adathick
26: Missing examples :
26: Differing graphical output : 08 19
26: Missing stdout :
26: Differing stdout :
26: ocaml
26: Missing examples :
26: Differing graphical output : 08 19 33
26: Missing stdout :
26: Differing stdout :
26: WARNING: Some graphical or stdout results were different
5/5 Test #26: examples_compare ................. Passed 2.33 sec
The following tests passed:
examples_ocaml
examples_c
examples_cxx
examples_ada
examples_compare
100% tests passed, 0 tests failed out of 5
Prior to this fix, the Missing stdout lines were filled with all
example numbers.
M plplot_test/test_diff.sh.in
commit a8873275168bccd57676457180ff810331f5278b
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Tue Apr 21 12:57:39 2015 -0700
Build system: Tcl-related bindings: use generator expressions instead of LOCATION properties
This is a fairly large step in the war against using LOCATION properties
to get into conformance with setting CMP0026 to new.
Since generator expressions only work at generate-time (which
apparently comes later than configure-time), can no longer use
configure-time cmake logic to concatenate the various pkgIndex.tcl
files together into one joint pkgIndex.tcl file that is installed.
Instead, use a custom_command/custom_target pair to do that
concatenation at build time using the concatenate_pkgIndex.tcl target.
Also drop some leftovers from the "d" suffix era in the configured
pkgIndex.tcl files.
Tested by Alan W. Irwin <airwin@users.sourceforge.net> on Linux
by building the concatenate_pkgIndex.tcl target in the build tree
and inspecting the configured results that were implemented with
generator expressions.
M bindings/CMakeLists.txt
M bindings/tcl/CMakeLists.txt
M bindings/tcl/pkgIndex.tcl.in
M bindings/tk-x-plat/CMakeLists.txt
M bindings/tk-x-plat/pkgIndex.tcl.in
M bindings/tk/CMakeLists.txt
M bindings/tk/pkgIndex.tcl.in
commit 7d93dba80e74445bc9055800d3a38a42d1be7300
Author: Arjen Markus <arjen.markus@deltares.nl>
Date: Tue Apr 21 21:59:07 2015 +0200
Extend the procedure to determine if we are in the build tree so that it works for MinGW/MSYS - the name of the directory is case-insenstive in this environment and in some versions forward slashes are used rather than backward slashes.
Tested on MinGW/MSYS by Arjen Markus, but should also work on other Windows environments, as the CMake logic uses the WIN32_OR_CYGWIN variable and the C code uses a macro IGNORECASE, in stead of a compiler-specific macro.
M src/CMakeLists.txt
M src/plcore.c
commit 8ffab0b2920c468acae09e0d31ed5925bc456405
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Mon Apr 20 18:19:53 2015 -0700
libplplot: plInBuildTree: bugfix: For windows check both possible directory separators
Previously the logic assumed the directory separator was '\\' on
Windows, but with recent versions of MSYS this may have changed so
now for Windows platforms check both the '\\' and '/' separators.
Tested by Alan W. Irwin <airwin@users.sourceforge.net> on both Linux
and MinGW/MSYS/Wine by using the -debug option to make sure the
build-tree drivers directory is properly found in both cases, i.e.,
plInBuildTree was correctly identifying that the code was being run
from the build tree on both platforms. Note, this test worked well
before because I had a moderately old MSYS version installed that only
uses the '\\' directory separator in this context. So the definitive
test will be by Arjen where plInBuildTree was not correctly
identifying that the code was being run from the build tree so that
the build-directory drivers directory could not be found on his more
modern MSYS system.
M src/plcore.c
commit c44f63607d167eba450020f768ad7dfb259c6cfb
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Mon Apr 20 17:22:42 2015 -0700
libplplot: Fix bug in -debug
The -debug option did not work previously for plInitDispatchTable
because plparseopt called pllib_init before processing command-line
arguments. I assume some of the command-line options do require
pllib_init to be called first, but -debug is not one of them.
Therefore, to solve this bug I process the -debug option before
the call to pllib_init, and process the rest of the options
after that call. The plparseopts logic is notorious for being fragile,
and I am near or beyond the level of my C expertise so I thoroughly
tested this change (see below) before committing it.
Tested by Alan W. Irwin <airwin@users.sourceforge.net> on Linux
in the build tree using
valgrind examples/c/x00c -debug -dev psc -o test.psc
(which gave perfect results), and using
make -j4 test_noninteractive >& test_noninteractive.out
make -j4 test_interactive >& test_interactive.out
(neither of which had any obvious errors).
M src/plargs.c
commit d8d999940dba17164878101cb40e1f2fd12718d7
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sat Apr 18 19:23:40 2015 -0700
Traditional build system for installed examples: Improve linking for non-shared case
The problem is that libplplot is a mixed C and C++ library for the
nondynamic or static case (because of including C++ drivers code in
the library for those cases). Therefore, for the traditional build
system for the installed examples, when gcc is used to link a C
example a specific link to libstdc++.so has to be provided.
The code in src/CMakeLists.txt to find that library is really ugly,
works only for g++ and no other C++ compiler, and depends on g++
internals which apparently have now changed. (Uses colon-separation
of libraries rather than semicolon separation.) And figuring out how
to deal with that colon separation when there are drive letters is a
real pain to deal with.
So I have completely abandoned that method and instead simply use the
C++ compiler (which automatically links in libstdc++) for the
non-shared case when building C examples using the traditional build
system. For some reason I had to change how plexit was declared in
examples/c/x08c.c to get this idea to work.
Tested by Alan W. Irwin <airwin@users.sourceforge.net> on Linux
using
scripts/comprehensive_test.sh \
--do_ctest no \
--do_test_interactive no \
--do_test_build_tree no \
--do_test_install_tree no
M examples/c/CMakeLists.txt
M examples/c/x08c.c
M examples/tk/CMakeLists.txt
M src/CMakeLists.txt
commit d512b0e7520d833e3aa3587abe94e69392a0f347
Author: Andrew Ross <andrewross@users.sourceforge.net>
Date: Sat Apr 18 21:59:31 2015 +0100
Make pl_MenuStr and pl_DevName const char * not char * since these strings are never changed. This fixes a gcc compiler warning about assigning strings to char * for the static library case.
M include/disptab.h
commit afda9cc3544d544f83cfb6e81ab8c4a2d422dbca
Author: Andrew Ross <andrewross@users.sourceforge.net>
Date: Thu Mar 19 09:27:52 2015 +0000
Fix bug where plD_tidy_* routine wasn't closed if the xwin / xcairo drivers were exited using the close button.
M src/plcore.c
commit c689ff3b60bf0f2e8f3381dc2d24ad8d1d49a2d1
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Sat Apr 18 10:20:29 2015 -0700
Comprehensive testing script: restore correct Windows PATH manipulation
The Windows PATH manipulation should put the dll subdirectory in the
build tree in the PATH for build-tree tests for BOTH the shared and
nondynamic cases. Due to a misunderstanding when rewriting this logic
after the 5.11.0 release, I changed this logic to forget about the
nondynamic case, and Arjen's recent Cygwin test failed in the
nondynamic case because of that. This commit restores that correct
behaviour.
M scripts/comprehensive_test.sh
commit 02be839d66ef7217d32503e6abf7d80f6988cc45
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Fri Apr 17 12:45:22 2015 -0700
Comprehensive testing script: Improve OS detection by using OSTYPE variable
M scripts/comprehensive_test.sh
commit 8972f92d037bd260ba9dea5da25ed662a53272c0
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Thu Apr 16 13:56:58 2015 -0700
Build system: Adopt a minimum CMake version of 3.0.2 everywhere
In order for our current build system to work, we have had to
explicitly set several CMake policies to old, but using such
deprecated features of CMake is not ideal so I plan to make
changes in our build system so that it works with new policies.
The problem is, however, that those planned changes use CMake
functionality that is only available for CMake version 3. Therefore,
as a prerequisite for these planned changes, in this commit I have
updated the minimum allowed version of CMake to 3.0.2 everywhere in
our source tree that the cmake_minimum_required command is used.
M CMakeLists.txt
M cmake/modules/language_support.cmake
M cmake/test_ada/CMakeLists.txt
M examples/CMakeLists.txt
commit 13a3baa3f7ade433d3b532017954f586d3fd80c4
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Tue Apr 14 12:52:58 2015 -0700
Comprehensive testing bug-fix for required PATH manipulations on
Windows platforms
This script uses the following PATH manipulations on Windows
platforms.
1. Put the dll subdirectory of the build tree on the PATH for _just_
the build-tree parts of the testing.
2. Put the installed bin directory and drivers directory on the PATH
for _just_ the install-tree parts of the testing.
As discussed on list prior to the 5.11.0 release, these required
Windows PATH manipulations were only being done on MinGW and
MinGW/MSYS which lead to script errors on Cygwin (unless the PATH was
brute-forced prior to running the script in a way that mixed
build-tree and install-tree results together).
As far as I know, these PATH manipulations should be done on all
Windows platforms. Therefore, this bug fix detects Windows platforms
in a much more general way now (by looking for the "exe" suffix for
the full pathname of the bash command), and this should solve
the Cygwin issue (i.e., the brute-force PATH manipulations should no
longer be required) and similar issues that would have occurred for MSVC
once that platform was comprehensively tested with this script.
Tested by Alan W. Irwin <airwin@users.sourceforge.net> on Linux
using
scripts/comprehensive_test.sh \
--do_nondynamic no \
--do_static no \
--do_ctest no \
--do_test_interactive no \
--do_test_traditional_install_tree no
i.e, run the test_noninteractive target in the build tree and
the non-traditional installed examples tree for the shared library
case only.
Note, this test merely checks that the script still works on Linux,
but the definitive test of this bug fix will require running this
script on Cygwin and other Windows platforms.
M scripts/comprehensive_test.sh
commit af95620793831915205f6f6b15226b14f27688af
Author: Alan W. Irwin <airwin@users.sourceforge.net>
Date: Mon Apr 13 12:19:09 2015 -0700
Prepare for next release cycle
Last update for README.Release_Manager_Cookbook to reflect what
was done for the release of 5.11.0.
Save 5.11.0 release notes in OLD-README.release.
Prepare placeholder (README.release) for release notes for the next
release.
M OLD-README.release
M README.Release_Manager_Cookbook
M README.release
Download Latest Version
plplot-5.15.0.tar.gz (15.4 MB)
