plplot-devel

[Plplot-devel] Building fails with Lua version > 5.1

[Atri <bad...@ai...>]

Hi!

I package plplot (version 5.9.9) for the Linux distribution openSUSE
[1]. I have run into a problem with building plplot for openSUSE's
factory (eventually this will become the next openSUSE release): the
version of lua in factory is 5.2 [2], but plplot's lua bindings won't
build with lua > 5.1. For now I have just disabled the lua bindings for
factory. It seems to me that the cmake module for building the lua
bindings check explicitly for versions of lua = 5.0 and 5.1. It might be
useful to adapt this cmake module to accept any lua version >= 5.0. I
have really little knowledge about cmake, but could help out with build
testing etc. as the need be.

Thanks a lot!

-- 
Atri

[1]
https://build.opensuse.org/package/show?package=plplot&project=science
[2] http://www.lua.org/versions.html#5.2

Re: [Plplot-devel] Building fails with Lua version > 5.1

[Alan W. I. <ir...@be...>]

On 2012-02-11 19:53+0530 Atri wrote:

> Hi!
>
> I package plplot (version 5.9.9) for the Linux distribution openSUSE
> [1]. I have run into a problem with building plplot for openSUSE's
> factory (eventually this will become the next openSUSE release): the
> version of lua in factory is 5.2 [2], but plplot's lua bindings won't
> build with lua > 5.1. For now I have just disabled the lua bindings for
> factory. It seems to me that the cmake module for building the lua
> bindings check explicitly for versions of lua = 5.0 and 5.1. It might be
> useful to adapt this cmake module to accept any lua version >= 5.0. I
> have really little knowledge about cmake, but could help out with build
> testing etc. as the need be.

Hi Atri:

Lua-5.2 was released only in December so it will take a while for that
to propagate to most Linux distros.  For example, Debian is quite
conservative and only supports Lua-5.0 even in the unstable version
of Debian for some reason.

So the bad news is you will be somewhat on your own (i.e., the only
potential tester of Lua-5.2 with PLplot that I am aware of) for a
while, but the good news is the CMake build system should be really
easy to modify so that you can find and attempt to build and test
PLplot using Lua 5.2.

If you are interested further, here is what I suggest you do.

Copy the system version of
FindLua51.cmake (usually found in
/usr/share/cmake-2.8/Modules) to cmake/modules.

That copy operation assures that the PLplot system will use the
version of FindLua51.cmake in cmake/modules rather than
the system one.

Then edit the PLplot version as follows:

Wherever there is a reference to Lua 5.1 put in additional 5.2 choices
first.

That is:

PATH_SUFFIXES include/lua51 include/lua5.1 include/lua include

==>

PATH_SUFFIXES include/lua52 include/lua5.2 include/lua51 include/lua5.1 include/lua include


NAMES lua51 lua5.1 lua-5.1 lua

==>

NAMES lua52 lua5.2 lua-5.2 lua51 lua5.1 lua-5.1 lua

Leave

FIND_PACKAGE_HANDLE_STANDARD_ARGS(Lua51 ...

as is since the 51 there refers to the name of the find module itself
which for now should continue to be cmake/modules/FindLua51.cmake.

I think the above two changes should be all the changes you need so
that the PLplot CMake-based build system will find the lua-5.2 library
and header files.

At that point, you will be in the interesting position of being able
test whether PLplot will work without further changes with lua-5.2. 
As far as I know, PLplot makes no distinctions between 5.0 and 5.1 so
I think 5.2 has a reasonable chance to work, but we won't know that
until you try it.

If your PLplot lua-5.2 build goes okay (with -DBUILD_TEST=ON cmake
option) with the above copy and edit of FindLua51.cmake, then please
also try run-time testing of our lua bindings using the following
command:

make test_lua_psc

which runs all our lua examples with -dev psc.

If that works without issues (i.e., the lua results in the examples
subdirectory of the build tree render okay with a PostScript viewer
like gv) I would be happy to copy your modified version of
FindLua51.cmake to PLplot so all PLplot/Lua-5.2 users can benefit.

Good luck, and let us know how it goes.

Also, thank you for supporting PLplot by packaging it
for openSUSE.

Alan

__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state
implementation for stellar interiors (freeeos.sf.net); the Time
Ephemerides project (timeephem.sf.net); PLplot scientific plotting
software package (plplot.sf.net); the libLASi project
(unifont.org/lasi); the Loads of Linux Links project (loll.sf.net);
and the Linux Brochure Project (lbproject.sf.net).
__________________________

Linux-powered Science
__________________________

Re: [Plplot-devel] Building fails with Lua version > 5.1

[Alan W. I. <ir...@be...>]

On 2012-02-11 14:57-0800 Alan W. Irwin wrote:

> Lua-5.2 was released only in December so it will take a while for that
> to propagate to most Linux distros.  For example, Debian is quite
> conservative and only supports Lua-5.0 even in the unstable version
> of Debian for some reason.

The first sentence is right which still implies it will be a while
before very many PLplot users can help with 5.2 testing, but I would
like to correct the Debian misinformation in the rest of that
paragraph.

Basically, I was confused by a change in lua package naming schemes so
I missed the more modern lua packages for Debian in that comment.

To correct that comment, the lua50 and lua5.1 (note the changed naming
convention) Debian packages are in Debian stable, testing, and
unstable. The lua5.2 Debian package is only in Debian experimental
now, and I expect it will be after the next release of Debian before
it becomes part of Debian unstable and eventually gets propagated from
there (if it passes all tests) to Debian testing.

I do plan to upgrade from Debian stable to testing in late March, but
even then I don't expect to have access to a packaged version of lua
5.2 from the above results on the packaging of lua for Debian. 
Therefore, I hope the three lines of changes I have suggested to Atri
for our build system so that it can find lua 5.2 will allow him to
test whether PLplot actually builds and runs OK with the 5.2 version
of lua.

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state
implementation for stellar interiors (freeeos.sf.net); the Time
Ephemerides project (timeephem.sf.net); PLplot scientific plotting
software package (plplot.sf.net); the libLASi project
(unifont.org/lasi); the Loads of Linux Links project (loll.sf.net);
and the Linux Brochure Project (lbproject.sf.net).
__________________________

Linux-powered Science
__________________________

Re: [Plplot-devel] Building fails with Lua version > 5.1

[Andrew R. <and...@us...>]

On Sat, Feb 11, 2012 at 02:57:44PM -0800, Alan Irwin wrote:
> On 2012-02-11 19:53+0530 Atri wrote:
> 
> > Hi!
> >
> > I package plplot (version 5.9.9) for the Linux distribution openSUSE
> > [1]. I have run into a problem with building plplot for openSUSE's
> > factory (eventually this will become the next openSUSE release): the
> > version of lua in factory is 5.2 [2], but plplot's lua bindings won't
> > build with lua > 5.1. For now I have just disabled the lua bindings for
> > factory. It seems to me that the cmake module for building the lua
> > bindings check explicitly for versions of lua = 5.0 and 5.1. It might be
> > useful to adapt this cmake module to accept any lua version >= 5.0. I
> > have really little knowledge about cmake, but could help out with build
> > testing etc. as the need be.
> 
> Hi Atri:
> 
> Lua-5.2 was released only in December so it will take a while for that
> to propagate to most Linux distros.  For example, Debian is quite
> conservative and only supports Lua-5.0 even in the unstable version
> of Debian for some reason.
> 
> So the bad news is you will be somewhat on your own (i.e., the only
> potential tester of Lua-5.2 with PLplot that I am aware of) for a
> while, but the good news is the CMake build system should be really
> easy to modify so that you can find and attempt to build and test
> PLplot using Lua 5.2.

Hi Atri,

That is not entirely true. Lua5.1 has been in debian for some time (it
is present in lenny (oldstable) which was released several years ago 
now. 

Lua5.2 (or at least a RC of the final release) is available in
Debian experimental, so there is some work going on there. Anyway, 
the point still stands that the core developers don't currently have
easy access to lua 5.2, so you help in testing will be invaluable.

Anyway, in the case of plplot there are two external issues. One is
the cmake support (which you have already encountered and which Alan
has offered some advice on). The other is that the lua bindings are
generated by swig. A quick google search suggests there are issues
with swig and lua 5.2 

http://lua-users.org/lists/lua-l/2011-12/msg00803.html

but it would be good to confirm / refute this with plplot. There is
a patch for swig on sourceforge which you could test

http://sourceforge.net/tracker/index.php?func=detail&aid=3480366&group_id=1645&atid=301645

Do let us know how you get on.

Regards

Andrew

Re: [Plplot-devel] Building fails with Lua version > 5.1

[Atri <bad...@ai...>]

On Sat, 2012-02-11 at 14:57 -0800, Alan W. Irwin wrote:
> On 2012-02-11 19:53+0530 Atri wrote:
> 
> > Hi!
> >
> > I package plplot (version 5.9.9) for the Linux distribution openSUSE
> > [1]. I have run into a problem with building plplot for openSUSE's
> > factory (eventually this will become the next openSUSE release): the
> > version of lua in factory is 5.2 [2], but plplot's lua bindings won't
> > build with lua > 5.1. For now I have just disabled the lua bindings for
> > factory. It seems to me that the cmake module for building the lua
> > bindings check explicitly for versions of lua = 5.0 and 5.1. It might be
> > useful to adapt this cmake module to accept any lua version >= 5.0. I
> > have really little knowledge about cmake, but could help out with build
> > testing etc. as the need be.
> 
> Hi Atri:
> 
> Lua-5.2 was released only in December so it will take a while for that
> to propagate to most Linux distros.  For example, Debian is quite
> conservative and only supports Lua-5.0 even in the unstable version
> of Debian for some reason.
> 
> So the bad news is you will be somewhat on your own (i.e., the only
> potential tester of Lua-5.2 with PLplot that I am aware of) for a
> while, but the good news is the CMake build system should be really
> easy to modify so that you can find and attempt to build and test
> PLplot using Lua 5.2.
> 
> If you are interested further, here is what I suggest you do.
> 
> Copy the system version of
> FindLua51.cmake (usually found in
> /usr/share/cmake-2.8/Modules) to cmake/modules.
> 
> That copy operation assures that the PLplot system will use the
> version of FindLua51.cmake in cmake/modules rather than
> the system one.
> 
> Then edit the PLplot version as follows:
> 
> Wherever there is a reference to Lua 5.1 put in additional 5.2 choices
> first.
> 
> That is:
> 
> PATH_SUFFIXES include/lua51 include/lua5.1 include/lua include
> 
> ==>
> 
> PATH_SUFFIXES include/lua52 include/lua5.2 include/lua51 include/lua5.1 include/lua include
> 
> 
> NAMES lua51 lua5.1 lua-5.1 lua
> 
> ==>
> 
> NAMES lua52 lua5.2 lua-5.2 lua51 lua5.1 lua-5.1 lua
> 
> Leave
> 
> FIND_PACKAGE_HANDLE_STANDARD_ARGS(Lua51 ...
> 
> as is since the 51 there refers to the name of the find module itself
> which for now should continue to be cmake/modules/FindLua51.cmake.
> 
> I think the above two changes should be all the changes you need so
> that the PLplot CMake-based build system will find the lua-5.2 library
> and header files.
> 


I tried to do that following your suggestions. So I now bundled up
FindLua51.cmake in the plplot source directory and tried to do a build
after the suggested modifications. So the file FindLua51.cmake in the
directory cmake/modules has this:-

===================================
FIND_PATH(LUA_INCLUDE_DIR lua.h
  HINTS
  $ENV{LUA_DIR}
  PATH_SUFFIXES include/lua52 include/lua5.2 include/lua51
include/lua5.1 include/lua include
  PATHS
  ~/Library/Frameworks
  /Library/Frameworks
  /usr/local
  /usr
  /sw # Fink
  /opt/local # DarwinPorts
  /opt/csw # Blastwave
  /opt
)

FIND_LIBRARY(LUA_LIBRARY 
  NAMES lua52 lua5.2 lua-5.2 lua51 lua5.1 lua-5.1 lua
  HINTS
  $ENV{LUA_DIR}
  PATH_SUFFIXES lib64 lib
  PATHS
  ~/Library/Frameworks
  /Library/Frameworks
  /usr/local
  /usr
  /sw
  /opt/local
  /opt/csw
  /opt
)
===============================


However the build still fails with the following messages:-

===============================
CMake Error at cmake/modules/FindLua51.cmake:69 (INCLUDE):
  include could not find load file:

    /home/abuild/rpmbuild/BUILD/plplot-5.9.9/cmake/modules/FindPackageHandleStandardArgs.cmake
Call Stack (most recent call first):
  cmake/modules/lua.cmake:63 (find_package)
  cmake/modules/plplot.cmake:475 (include)
  CMakeLists.txt:91 (include)


-- Found Lua51: /usr/lib64/liblua.so;/usr/lib64/libm.so 
-- LUA_VERSION = 5.2
-- WARNING: Lua executable found but version not consistent with
library. Disabling Lua bindings
===============================

I am sure I am doing something wrong, but can't quite figure it out.


> Also, thank you for supporting PLplot by packaging it
> for openSUSE.
> 
> Alan
> 

Thanks for your extensive help.



-- 
Atri

Re: [Plplot-devel] Building fails with Lua version > 5.1

[Alan W. I. <ir...@be...>]

On 2012-02-12 18:26+0530 Atri wrote:

> On Sat, 2012-02-11 at 14:57 -0800, Alan W. Irwin wrote:

>> Wherever there is a reference to Lua 5.1 put in additional 5.2 choices
>> first.
>>
>> That is:
>>
>> PATH_SUFFIXES include/lua51 include/lua5.1 include/lua include
>>
>> ==>
>>
>> PATH_SUFFIXES include/lua52 include/lua5.2 include/lua51 include/lua5.1 include/lua include
>>
>>
>> NAMES lua51 lua5.1 lua-5.1 lua
>>
>> ==>
>>
>> NAMES lua52 lua5.2 lua-5.2 lua51 lua5.1 lua-5.1 lua
[...]
> I tried to do that following your suggestions. So I now bundled up
> FindLua51.cmake in the plplot source directory and tried to do a build
> after the suggested modifications [...]

> However the build still fails with the following messages:-
>
> ===============================
> CMake Error at cmake/modules/FindLua51.cmake:69 (INCLUDE):
>  include could not find load file:
>
>    /home/abuild/rpmbuild/BUILD/plplot-5.9.9/cmake/modules/FindPackageHandleStandardArgs.cmake
> Call Stack (most recent call first):
>  cmake/modules/lua.cmake:63 (find_package)
>  cmake/modules/plplot.cmake:475 (include)
>  CMakeLists.txt:91 (include)
>
>
> -- Found Lua51: /usr/lib64/liblua.so;/usr/lib64/libm.so
> -- LUA_VERSION = 5.2
> -- WARNING: Lua executable found but version not consistent with
> library. Disabling Lua bindings
> ===============================
>

Sorry, I completely missed one additional change in my directions.

Make the following change as well:

INCLUDE(${CMAKE_CURRENT_LIST_DIR}/FindPackageHandleStandardArgs.cmake)

==>

INCLUDE(FindPackageHandleStandardArgs)

Note, the second form must have ".cmake" dropped as indicated.

The first form looks only in cmake/modules for
FindPackageHandleStandardArgs.cmake, and since it is not there you got
the above error message.  The second form looks in cmake/modules
first, but if not there looks in the system location for
FindPackageHandleStandardArgs.cmake (which is what we want in this
case).

Hope that additional change allows our build system to find lua5.2,
but if not, I would be happy to help some more.

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state
implementation for stellar interiors (freeeos.sf.net); the Time
Ephemerides project (timeephem.sf.net); PLplot scientific plotting
software package (plplot.sf.net); the libLASi project
(unifont.org/lasi); the Loads of Linux Links project (loll.sf.net);
and the Linux Brochure Project (lbproject.sf.net).
__________________________

Linux-powered Science
__________________________

Re: [Plplot-devel] Building fails with Lua version > 5.1

[Atri <bad...@ai...>]

Hi!

Thanks for all the helpful suggestions. My first attempt at building
with lua 5.2 is outlined below.

On Sat, 2012-02-11 at 14:57 -0800, Alan W. Irwin wrote:
> Copy the system version of
> FindLua51.cmake (usually found in
> /usr/share/cmake-2.8/Modules) to cmake/modules.
> 
> That copy operation assures that the PLplot system will use the
> version of FindLua51.cmake in cmake/modules rather than
> the system one.
> 
> Then edit the PLplot version as follows:
> 
> Wherever there is a reference to Lua 5.1 put in additional 5.2 choices
> first.
> 
> That is:
> 
> PATH_SUFFIXES include/lua51 include/lua5.1 include/lua include
> 
> ==>
> 
> PATH_SUFFIXES include/lua52 include/lua5.2 include/lua51 include/lua5.1 include/lua include
> 
> 
> NAMES lua51 lua5.1 lua-5.1 lua
> 
> ==>
> 
> NAMES lua52 lua5.2 lua-5.2 lua51 lua5.1 lua-5.1 lua
> 
> Leave
> 
> FIND_PACKAGE_HANDLE_STANDARD_ARGS(Lua51 ...
> 
> as is since the 51 there refers to the name of the find module itself
> which for now should continue to be cmake/modules/FindLua51.cmake.
> 

In addition to these changes, I also had to do the following to the file
cmake/modules/lua.cmake

--- lua.cmake.orig	2011-10-13 06:13:01.000000000 +0530
+++ lua.cmake	2012-02-13 00:03:36.000000000 +0530
@@ -100,6 +100,8 @@
     set(LUA_VERSION_VALID ON)
   elseif(${LUA_VERSION} STREQUAL "5.1" AND LUA51_FOUND)
     set(LUA_VERSION_VALID ON)
+  elseif(${LUA_VERSION} STREQUAL "5.2" AND LUA51_FOUND)
+    set(LUA_VERSION_VALID ON)
   endif(${LUA_VERSION} STREQUAL "5.0" AND LUA50_FOUND)
 
   if(NOT LUA_VERSION_VALID)
---

After these modifications to the source directory, and using a modified
version of swig-2.0.4 with the following patch applied
http://sourceforge.net/tracker/index.php?func=detail&aid=3480366&group_id=1645&atid=301645
I got plplot along with its lua bindings to build successfully with lua
5.2. 


> If your PLplot lua-5.2 build goes okay (with -DBUILD_TEST=ON cmake
> option) with the above copy and edit of FindLua51.cmake, then please
> also try run-time testing of our lua bindings using the following
> command:
> 
> make test_lua_psc
> 
> which runs all our lua examples with -dev psc.
> 
> If that works without issues (i.e., the lua results in the examples
> subdirectory of the build tree render okay with a PostScript viewer
> like gv) I would be happy to copy your modified version of
> FindLua51.cmake to PLplot so all PLplot/Lua-5.2 users can benefit.
> __________________________
> Alan W. Irwin
> 

Now here is where the problem occurs. When attempting the 'make
test_lua_psc' the first example 'x01' seems to work fine. However, for
'x02',

---
Testing front-end lua
x01
x02
/usr/bin/lua: x02.lua:105: attempt to call field 'mod' (a nil value)
stack traceback:
	x02.lua:105: in function 'demo2'
	x02.lua:150: in main chunk
	[C]: in ?
make[3]: *** [examples/x01lua.psc] Error 1
make[2]: *** [examples/CMakeFiles/test_lua_psc.dir/all] Error 2
make[1]: *** [examples/CMakeFiles/test_lua_psc.dir/rule] Error 2
make: *** [test_lua_psc] Error 2
error: Bad exit status from /var/tmp/rpm-tmp.RoZaxq (%build)
---

... and there the build halts with an error. Any suggestions about this?

As a side-note, I also got plplot's lua bindings building for openSUSE
Factory (latest) by explicitly using lua < 5.2 which is also supported
for the foreseeable future :). It will be good, nontheless, if we fix
issues with 5.2 as well.

Bye and thanks again for all the help.

-- 
Atri

Re: [Plplot-devel] Building fails with Lua version > 5.1

[Alan W. I. <ir...@be...>]

On 2012-02-16 22:36+0530 Atri wrote:

> Hi!
>
> Thanks for all the helpful suggestions. My first attempt at building
> with lua 5.2 is outlined below.
>
> On Sat, 2012-02-11 at 14:57 -0800, Alan W. Irwin wrote:
>> Copy the system version of
>> FindLua51.cmake (usually found in
>> /usr/share/cmake-2.8/Modules) to cmake/modules.
>>
>> That copy operation assures that the PLplot system will use the
>> version of FindLua51.cmake in cmake/modules rather than
>> the system one.
>>
>> Then edit the PLplot version as follows:
>>
>> Wherever there is a reference to Lua 5.1 put in additional 5.2 choices
>> first.
>>
>> That is:
>>
>> PATH_SUFFIXES include/lua51 include/lua5.1 include/lua include
>>
>> ==>
>>
>> PATH_SUFFIXES include/lua52 include/lua5.2 include/lua51 include/lua5.1 include/lua include
>>
>>
>> NAMES lua51 lua5.1 lua-5.1 lua
>>
>> ==>
>>
>> NAMES lua52 lua5.2 lua-5.2 lua51 lua5.1 lua-5.1 lua
>>
>> Leave
>>
>> FIND_PACKAGE_HANDLE_STANDARD_ARGS(Lua51 ...
>>
>> as is since the 51 there refers to the name of the find module itself
>> which for now should continue to be cmake/modules/FindLua51.cmake.
>>
>
> In addition to these changes, I also had to do the following to the file
> cmake/modules/lua.cmake
>
> --- lua.cmake.orig	2011-10-13 06:13:01.000000000 +0530
> +++ lua.cmake	2012-02-13 00:03:36.000000000 +0530
> @@ -100,6 +100,8 @@
>     set(LUA_VERSION_VALID ON)
>   elseif(${LUA_VERSION} STREQUAL "5.1" AND LUA51_FOUND)
>     set(LUA_VERSION_VALID ON)
> +  elseif(${LUA_VERSION} STREQUAL "5.2" AND LUA51_FOUND)
> +    set(LUA_VERSION_VALID ON)
>   endif(${LUA_VERSION} STREQUAL "5.0" AND LUA50_FOUND)
>
>   if(NOT LUA_VERSION_VALID)
> ---
>
> After these modifications to the source directory, and using a modified
> version of swig-2.0.4 with the following patch applied
> http://sourceforge.net/tracker/index.php?func=detail&aid=3480366&group_id=1645&atid=301645
> I got plplot along with its lua bindings to build successfully with lua
> 5.2.

Good.

>
>
>> If your PLplot lua-5.2 build goes okay (with -DBUILD_TEST=ON cmake
>> option) with the above copy and edit of FindLua51.cmake, then please
>> also try run-time testing of our lua bindings using the following
>> command:
>>
>> make test_lua_psc
>>
>> which runs all our lua examples with -dev psc.
>>
>> If that works without issues (i.e., the lua results in the examples
>> subdirectory of the build tree render okay with a PostScript viewer
>> like gv) I would be happy to copy your modified version of
>> FindLua51.cmake to PLplot so all PLplot/Lua-5.2 users can benefit.
>> __________________________
>> Alan W. Irwin
>>
>
> Now here is where the problem occurs. When attempting the 'make
> test_lua_psc' the first example 'x01' seems to work fine. However, for
> 'x02',
>
> ---
> Testing front-end lua
> x01
> x02
> /usr/bin/lua: x02.lua:105: attempt to call field 'mod' (a nil value)
> stack traceback:
> 	x02.lua:105: in function 'demo2'
> 	x02.lua:150: in main chunk
> 	[C]: in ?
> make[3]: *** [examples/x01lua.psc] Error 1
> make[2]: *** [examples/CMakeFiles/test_lua_psc.dir/all] Error 2
> make[1]: *** [examples/CMakeFiles/test_lua_psc.dir/rule] Error 2
> make: *** [test_lua_psc] Error 2
> error: Bad exit status from /var/tmp/rpm-tmp.RoZaxq (%build)
> ---
>
> ... and there the build halts with an error. Any suggestions about this?

The fact that you got the lua-5.2 build to work and example 1 (which
actually is pretty complicated) to run shows, I think, that
you are down to really specific issues as opposed to "nothing works".
So the prospects for PLplot to completely support lua-5.2 look extremely
promising.

The above "make test_lua_psc" command sets up being able to run
individual lua examples by hand.

Here is how you do that:

In the build directory....
cd examples/lua
lua x01.lua -dev psc -o test.ps

Does test.ps generated this way give a good-looking result using
the gv viewer of PostScript files?

If so, then I would try example 2 which should generate the above error
message

The offending line in question for example 2 is the innocuous looking

h = (360/10) * math.mod(i, 10)

I checked, and example 1 uses the math library as well so access to
that lua library appears not to be an issue.

So it appears that math.mod simply does not exist for 5.2. In fact,
the 5.2 documentation doesn't refer to it and the 5.1 documentation
http://www.lua.org/manual/5.1/manual.html says "Function math.mod was
renamed math.fmod".  Apparently my own 5.1 version of Lua only works
because of compatibility mode.

If you look up the math functions in lua in the above documentation
thy simply appear to be front ends for the C library version, e.g.,
math.fmod corresponds to the C library floating-point fmod function, and I presume
lua takes care of all the casting from integers for the above case.

So one possibility to deal with this is simply to replace all math.mod
in the several examples where that occurs with math.fmod.  However,
another alternative is to use the lua % operator (which takes the
remainder relative to minus infinity rather than zero as in the
math.fmod case).  If we are always dealing with positive values using
the % operator might be a better alternative than math.fmod.

Anyhow, let me test the math.fmod and % possibilities for all our lua
examples in the next little while, and I will get back to you.

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state
implementation for stellar interiors (freeeos.sf.net); the Time
Ephemerides project (timeephem.sf.net); PLplot scientific plotting
software package (plplot.sf.net); the libLASi project
(unifont.org/lasi); the Loads of Linux Links project (loll.sf.net);
and the Linux Brochure Project (lbproject.sf.net).
__________________________

Linux-powered Science
__________________________

Re: [Plplot-devel] Building fails with Lua version > 5.1

[Alan W. I. <ir...@be...>]

On 2012-02-16 11:46-0800 Alan W. Irwin wrote:

> [....]So one possibility to deal with this is simply to replace all math.mod
> in the several examples where that occurs with math.fmod.  However,
> another alternative is to use the lua % operator (which takes the
> remainder relative to minus infinity rather than zero as in the
> math.fmod case).  If we are always dealing with positive values using
> the % operator might be a better alternative than math.fmod.
>
> Anyhow, let me test the math.fmod and % possibilities for all our lua
> examples in the next little while, and I will get back to you.

In the end (revision 12167) I decided to change to the % operator and
the results passed the test_diff_psc test (i.e., were the same as
the corresponding C examples) for my Lua-5.1 platform.

So Atri, please try the test_lua_psc target again with this revision
from svn trunk. I hope you will be able to get a lot further into the
examples or maybe even complete them!

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state
implementation for stellar interiors (freeeos.sf.net); the Time
Ephemerides project (timeephem.sf.net); PLplot scientific plotting
software package (plplot.sf.net); the libLASi project
(unifont.org/lasi); the Loads of Linux Links project (loll.sf.net);
and the Linux Brochure Project (lbproject.sf.net).
__________________________

Linux-powered Science
__________________________

Re: [Plplot-devel] Building fails with Lua version > 5.1

[Atri <bad...@ai...>]

On Thu, 2012-02-16 at 12:31 -0800, Alan W. Irwin wrote:
> On 2012-02-16 11:46-0800 Alan W. Irwin wrote:
> 
> > [....]So one possibility to deal with this is simply to replace all math.mod
> > in the several examples where that occurs with math.fmod.  However,
> > another alternative is to use the lua % operator (which takes the
> > remainder relative to minus infinity rather than zero as in the
> > math.fmod case).  If we are always dealing with positive values using
> > the % operator might be a better alternative than math.fmod.
> >
> > Anyhow, let me test the math.fmod and % possibilities for all our lua
> > examples in the next little while, and I will get back to you.
> 
> In the end (revision 12167) I decided to change to the % operator and
> the results passed the test_diff_psc test (i.e., were the same as
> the corresponding C examples) for my Lua-5.1 platform.
> 
> So Atri, please try the test_lua_psc target again with this revision
> from svn trunk. I hope you will be able to get a lot further into the
> examples or maybe even complete them!
> 
> Alan
> __________________________

Hi!
For the build to be successful one needs to make another small change to
the file cmake/modules/lua.cmake, which is as follows:-

Index: plplot-5.9.9/cmake/modules/lua.cmake
===================================================================
--- plplot-5.9.9.orig/cmake/modules/lua.cmake
+++ plplot-5.9.9/cmake/modules/lua.cmake
@@ -115,7 +115,7 @@ if(ENABLE_lua)
   # Unless some better convention comes along, follow what Debian does
for
   # install location of Lua wrapper shared object.
   if(LUA51_FOUND)
-    set(LUA_DIR ${LIB_DIR}/lua/5.1/plplot)
+    set(LUA_DIR ${LIB_DIR}/lua/${LUA_VERSION}/plplot)
   else(LUA51_FOUND)
     set(LUA_DIR ${LIB_DIR}/lua/50/plplot)
   endif(LUA51_FOUND)
===================================================================

This patch ensures that if, for instance, lua = 5.2 (and indeed anything
greater than 5.0) is detected the corresponding plplot bindings are
installed in the directory $LIB_DIR/lua/5.2/plplot and so on. Without
this patch plplot bindings are installed in $LIB_DIR/lua/5.1/plplot even
if lua = 5.2 is detected in the system.

Finally, with these modifications, and the examples corrected as in the
latest revision, the 'make test_lua_psc' passes wonderfully, i.e. not
only does it pass, but it creates ps files of plots that look fine. I
found that the ps files created using the c++ examples and using lua 5.2
looked identical. I interpret that to mean plplot works just great with
lua 5.2 as well \o/

Do you want me to submit the ps files or something so you could verify?

Thanks a lot for all the really nice help. Please let me know if there
is something more I need to do about this.

Bye

-- 
Atri

Re: [Plplot-devel] Building fails with Lua version > 5.1

[Alan W. I. <ir...@be...>]

On 2012-02-17 23:52+0530 Atri wrote:

> Finally, with these modifications, and the examples corrected as in the
> latest revision, the 'make test_lua_psc' passes wonderfully, i.e. not
> only does it pass, but it creates ps files of plots that look fine. I
> found that the ps files created using the c++ examples and using lua 5.2
> looked identical. I interpret that to mean plplot works just great with
> lua 5.2 as well \o/
>
> Do you want me to submit the ps files or something so you could verify?

Hi Atri:

Please run the "test_diff_psc" target to verify in detail what you
have visually observed above.  What this test does is automatically
compare PostScript results from C standard examples with those same
examples implemented in all our supported languages (including lua).

Here are the lua-5.1 results of that test on my system:

lua
   Missing examples            :  00
   Differing postscript output :
   Missing stdout              :
   Differing stdout            :

which shows standard example 00 missing for lua, but for all other
30+ standard examples, lua and C results are the same.

In fact, I would recommend running the test_diff_psc target or better
yet the test_noninteractive target (which depends on test_diff_psc and
also a few other smaller non-interactive test targets) for anyone
interested in comprehensively checking their PLplot build.

>
> Thanks a lot for all the really nice help. Please let me know if there
> is something more I need to do about this.

You are welcome, and my thanks to you as well for pioneering lua 5.2
for us.

To complete this work I plan to update our build system to incorporate
all the changes you have shown are required to get lua 5.2 to work,
(with the added feature that I will use the lua -v command to
determine which version of lua is installed so we don't have to keep
maintaining version number lists).  Once that works for me here with
lua-5.1, I will ask you to do one further test of that unpatched updated build
system there to make sure it continues to work for lua-5.2.

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state
implementation for stellar interiors (freeeos.sf.net); the Time
Ephemerides project (timeephem.sf.net); PLplot scientific plotting
software package (plplot.sf.net); the libLASi project
(unifont.org/lasi); the Loads of Linux Links project (loll.sf.net);
and the Linux Brochure Project (lbproject.sf.net).
__________________________

Linux-powered Science
__________________________

Re: [Plplot-devel] Building fails with Lua version > 5.1

[Alan W. I. <ir...@be...>]

Hi Atri:

On 2012-02-17 12:18-0800 Alan W. Irwin wrote:

> To complete this work I plan to update our build system to incorporate
> all the changes you have shown are required to get lua 5.2 to work,
> (with the added feature that I will use the lua -v command to
> determine which version of lua is installed so we don't have to keep
> maintaining version number lists).

DONE as of revision 12168.  With this change I get good results for
the test_diff_psc target both for the build tree and CMake-based
installed examples tree for lua-5.1.

> Once that works for me here with
> lua-5.1, I will ask you to do one further test of that unpatched updated build
> system there to make sure it continues to work for lua-5.2.

Consider yourself so asked... :-)

Note, I have done the build system changes independently of the
details you presented simply by debugging for 5.1 what didn't work for
my new generalized support for lua-5.x.  So there might be some issues
left for 5.2 (and also 5.0 if anybody still has access to that), but I
hope not.

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state
implementation for stellar interiors (freeeos.sf.net); the Time
Ephemerides project (timeephem.sf.net); PLplot scientific plotting
software package (plplot.sf.net); the libLASi project
(unifont.org/lasi); the Loads of Linux Links project (loll.sf.net);
and the Linux Brochure Project (lbproject.sf.net).
__________________________

Linux-powered Science
__________________________

Re: [Plplot-devel] Building fails with Lua version > 5.1

[Atri <bad...@ai...>]

On Fri, 2012-02-17 at 12:18 -0800, Alan W. Irwin wrote:
> On 2012-02-17 23:52+0530 Atri wrote:
> 
> > Finally, with these modifications, and the examples corrected as in the
> > latest revision, the 'make test_lua_psc' passes wonderfully, i.e. not
> > only does it pass, but it creates ps files of plots that look fine. I
> > found that the ps files created using the c++ examples and using lua 5.2
> > looked identical. I interpret that to mean plplot works just great with
> > lua 5.2 as well \o/
> >
> > Do you want me to submit the ps files or something so you could verify?
> 
> Hi Atri:
> 
> Please run the "test_diff_psc" target to verify in detail what you
> have visually observed above.  What this test does is automatically
> compare PostScript results from C standard examples with those same
> examples implemented in all our supported languages (including lua).
> 
> Here are the lua-5.1 results of that test on my system:
> 
> lua
>    Missing examples            :  00
>    Differing postscript output :
>    Missing stdout              :
>    Differing stdout            :
> 
> which shows standard example 00 missing for lua, but for all other
> 30+ standard examples, lua and C results are the same.
> 
> In fact, I would recommend running the test_diff_psc target or better
> yet the test_noninteractive target (which depends on test_diff_psc and
> also a few other smaller non-interactive test targets) for anyone
> interested in comprehensively checking their PLplot build.

Ok, seems I jumped the gun there. Example 21 from latest svn seems to be
creating some problems. Here is the result of test_diff.sh

===============================================================
lua
  Missing examples            : 
  Differing postscript output :  21
  Missing stdout              : 
  Differing stdout            : 
WARNING: Some PostScript or stdout results were different
===============================================================

If instead I use math.fmod instead of % to replace all occurrences of
math.mod in the lua examples shipped with the 5.9.9 tarball, then I get
the expected results from test_diff.sh:-

===============================================================
lua
  Missing examples            : 
  Differing postscript output : 
  Missing stdout              : 
  Differing stdout            : 
===============================================================

Hope that helps.

> To complete this work I plan to update our build system to incorporate
> all the changes you have shown are required to get lua 5.2 to work,
> (with the added feature that I will use the lua -v command to
> determine which version of lua is installed so we don't have to keep
> maintaining version number lists).  Once that works for me here with
> lua-5.1, I will ask you to do one further test of that unpatched updated build
> system there to make sure it continues to work for lua-5.2.
> 
> Alan
> __________________________
> Alan W. Irwin
> 
> Astronomical research affiliation with Department of Physics and Astronomy,
> University of Victoria (astrowww.phys.uvic.ca).
> 
> Programming affiliations with the FreeEOS equation-of-state
> implementation for stellar interiors (freeeos.sf.net); the Time
> Ephemerides project (timeephem.sf.net); PLplot scientific plotting
> software package (plplot.sf.net); the libLASi project
> (unifont.org/lasi); the Loads of Linux Links project (loll.sf.net);
> and the Linux Brochure Project (lbproject.sf.net).
> __________________________
> 
> Linux-powered Science
> __________________________

Bye
-- 
Atri

Re: [Plplot-devel] Building fails with Lua version > 5.1

[Alan W. I. <ir...@be...>]

On 2012-02-18 04:42+0530 Atri wrote:

> On Fri, 2012-02-17 at 12:18 -0800, Alan W. Irwin wrote:
>> On 2012-02-17 23:52+0530 Atri wrote:
>>
>>> Finally, with these modifications, and the examples corrected as in the
>>> latest revision, the 'make test_lua_psc' passes wonderfully, i.e. not
>>> only does it pass, but it creates ps files of plots that look fine. I
>>> found that the ps files created using the c++ examples and using lua 5.2
>>> looked identical. I interpret that to mean plplot works just great with
>>> lua 5.2 as well \o/
>>>
>>> Do you want me to submit the ps files or something so you could verify?
>>
>> Hi Atri:
>>
>> Please run the "test_diff_psc" target to verify in detail what you
>> have visually observed above.  What this test does is automatically
>> compare PostScript results from C standard examples with those same
>> examples implemented in all our supported languages (including lua).
>>
>> Here are the lua-5.1 results of that test on my system:
>>
>> lua
>>    Missing examples            :  00
>>    Differing postscript output :
>>    Missing stdout              :
>>    Differing stdout            :
>>
>> which shows standard example 00 missing for lua, but for all other
>> 30+ standard examples, lua and C results are the same.
>>
>> In fact, I would recommend running the test_diff_psc target or better
>> yet the test_noninteractive target (which depends on test_diff_psc and
>> also a few other smaller non-interactive test targets) for anyone
>> interested in comprehensively checking their PLplot build.
>
> Ok, seems I jumped the gun there. Example 21 from latest svn seems to be
> creating some problems. Here is the result of test_diff.sh
>
> ===============================================================
> lua
>  Missing examples            :
>  Differing postscript output :  21
>  Missing stdout              :
>  Differing stdout            :
> WARNING: Some PostScript or stdout results were different
> ===============================================================

Hi Atri:

The first thing I noticed about the above result is there is no
complaint about missing example 00. So something appears to be wrong
with the conditions you used for this test.  My best guess
is you probably had a dirty build tree with some stale 5.9.9 results
mixed in with the trunk version results.

So could you please repeat this test starting from a fresh svn
checkout to an initially empty source tree, and then build starting
from an empty build tree?  That fresh checkout should capture my
recent build system changes so (with luck) there should be no need for
you to patch anything to get the build to work.

Once the unpatched build from trunk works for you, then please test
for run-time issues as above with the test_diff_psc target. If you
still get example 21 run-time issues as above, please try replacing
the % (just for that particular trunk example) with math.fmod.

The example 21 issue (if it still exists for a fresh start) may not be
with % (since that works perfectly for 5.1).  Instead, some other
trunk issue could be the cause of the problem, but replacing % with
math.fmod in the trunk example 21 should should help to answer that
question.

By the way, I ignored your 5.9.9 results since trunk has a number of
changes since 5.9.9, and any of those changes (or your stale build or
whatever it was that made the missing examples/lua/x00.lua go
unnoticed) could be the cause of the example 21 issue.  Also, getting
trunk working correctly is very important to the PLplot developers
since that version will become our next released version (whenever
that release occurs).

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state
implementation for stellar interiors (freeeos.sf.net); the Time
Ephemerides project (timeephem.sf.net); PLplot scientific plotting
software package (plplot.sf.net); the libLASi project
(unifont.org/lasi); the Loads of Linux Links project (loll.sf.net);
and the Linux Brochure Project (lbproject.sf.net).
__________________________

Linux-powered Science
__________________________

Re: [Plplot-devel] Building fails with Lua version > 5.1

[Atri <bad...@ai...>]

On Sat, 2012-02-18 at 11:37 -0800, Alan W. Irwin wrote:
> On 2012-02-18 04:42+0530 Atri wrote:
> 
> > Ok, seems I jumped the gun there. Example 21 from latest svn seems to be
> > creating some problems. Here is the result of test_diff.sh
> >
> > ===============================================================
> > lua
> >  Missing examples            :
> >  Differing postscript output :  21
> >  Missing stdout              :
> >  Differing stdout            :
> > WARNING: Some PostScript or stdout results were different
> > ===============================================================
> 
> Hi Atri:
> 
> The first thing I noticed about the above result is there is no
> complaint about missing example 00. So something appears to be wrong
> with the conditions you used for this test.  My best guess
> is you probably had a dirty build tree with some stale 5.9.9 results
> mixed in with the trunk version results.
> 
> So could you please repeat this test starting from a fresh svn
> checkout to an initially empty source tree, and then build starting
> from an empty build tree?  That fresh checkout should capture my
> recent build system changes so (with luck) there should be no need for
> you to patch anything to get the build to work.
> 

Hi!
Sorry for the long delay in between my last test and the present one. I
got the svn version to build on openSUSE:Factory (with Lua 5.2) finally.
I used for my build the svn trunk directory (at revision 12185) without
applying any patch except this following one, which is necessary to
avoid warnings with rpmlint during the build [1].

====================================================
diff -ur plplot-5.9.8.orig//drivers/wxwidgets.cpp
plplot-5.9.8//drivers/wxwidgets.cpp
--- plplot-5.9.8.orig//drivers/wxwidgets.cpp	2011-08-01
23:29:39.000000000 +0530
+++ plplot-5.9.8//drivers/wxwidgets.cpp	2011-08-02 01:06:57.323000001
+0530
@@ -231,7 +231,7 @@
             if ( ucs4[i] != (PLUNICODE) plplotEsc ) // a character to
display
             {
                 ucs4_to_utf8( ucs4[i], utf8 );
-                strncat( utf8_string, utf8, max_string_length );
+                strncat( utf8_string, utf8,
sizeof(utf8_string)-strlen(utf8_string)-1 );
                 i++;
                 continue;
             }
@@ -239,7 +239,7 @@
             if ( ucs4[i] == (PLUNICODE) plplotEsc ) // a escape
character to display
             {
                 ucs4_to_utf8( ucs4[i], utf8 );
-                strncat( utf8_string, utf8, max_string_length );
+                strncat( utf8_string, utf8,
sizeof(utf8_string)-strlen(utf8_string)-1);
                 i++;
                 continue;
             }
====================================================

After the building is done, I tested the package by running the
plplot_test.sh script followed by test_diff.sh test for lua and python.
The result is perfect, because the test reports no difference between
the psc plots obtained with c/cpp and those obtained with lua 5.2. This
is what the result says (I understand that the x00 example which was
missing from lua/python previously is now also available for these two
bindings):
============================================
python
  Missing examples            : 
  Differing postscript output : 
  Missing stdout              : 
  Differing stdout            : 
lua
  Missing examples            : 
  Differing postscript output : 
  Missing stdout              : 
  Differing stdout            : 
============================================

I can confirm the same result by manual inspection of the psc output
files, especially example 21 which was giving us problems earlier but
now looks exactly the same whether generated by a c/cpp code or by lua.
I believe we can now treat plplot at the current svn revision to be
perfectly compatible with lua 5.2 as well. Thanks a bunch for all of
your help with the tests, building etc. and for fixing everything too.

There is an unrelated problem about the GNAT library not being detected
by the build system, but I will report that in a separate mail. Please
let me know if you think any more tests could help with the lua 5.2
issue.

Thanks a lot and bye.
-- 
Atri

[1] The warnings without this patch says:-

I: Statement might be overflowing a buffer in strncat. Common mistake:
   BAD: strncat(buffer,charptr,sizeof(buffer)) is wrong, it takes the
left over size as 3rd argument
   GOOD: strncat(buffer,charptr,sizeof(buffer)-strlen(buffer)-1)
E: plplot
bufferoverflowstrncat /home/abuild/rpmbuild/BUILD/plplot/drivers/wxwidgets.cpp:253:64
E: plplot
bufferoverflowstrncat /home/abuild/rpmbuild/BUILD/plplot/drivers/wxwidgets.cpp:261:64

Re: [Plplot-devel] Building fails with Lua version > 5.1

[Andrew R. <and...@us...>]

On Wed, Apr 11, 2012 at 02:54:59AM +0530, Atri wrote:
> 
> Hi!
> Sorry for the long delay in between my last test and the present one. I
> got the svn version to build on openSUSE:Factory (with Lua 5.2) finally.
> I used for my build the svn trunk directory (at revision 12185) without
> applying any patch except this following one, which is necessary to
> avoid warnings with rpmlint during the build [1].
> 
> ====================================================
> diff -ur plplot-5.9.8.orig//drivers/wxwidgets.cpp
> plplot-5.9.8//drivers/wxwidgets.cpp
> --- plplot-5.9.8.orig//drivers/wxwidgets.cpp	2011-08-01
> 23:29:39.000000000 +0530
> +++ plplot-5.9.8//drivers/wxwidgets.cpp	2011-08-02 01:06:57.323000001
> +0530
> @@ -231,7 +231,7 @@
>              if ( ucs4[i] != (PLUNICODE) plplotEsc ) // a character to
> display
>              {
>                  ucs4_to_utf8( ucs4[i], utf8 );
> -                strncat( utf8_string, utf8, max_string_length );
> +                strncat( utf8_string, utf8,
> sizeof(utf8_string)-strlen(utf8_string)-1 );
>                  i++;
>                  continue;
>              }
> @@ -239,7 +239,7 @@
>              if ( ucs4[i] == (PLUNICODE) plplotEsc ) // a escape
> character to display
>              {
>                  ucs4_to_utf8( ucs4[i], utf8 );
> -                strncat( utf8_string, utf8, max_string_length );
> +                strncat( utf8_string, utf8,
> sizeof(utf8_string)-strlen(utf8_string)-1);
>                  i++;
>                  continue;
>              }
> ====================================================
> 
> After the building is done, I tested the package by running the
> plplot_test.sh script followed by test_diff.sh test for lua and python.
> The result is perfect, because the test reports no difference between
> the psc plots obtained with c/cpp and those obtained with lua 5.2. This
> is what the result says (I understand that the x00 example which was
> missing from lua/python previously is now also available for these two
> bindings):
> ============================================
> python
>   Missing examples            : 
>   Differing postscript output : 
>   Missing stdout              : 
>   Differing stdout            : 
> lua
>   Missing examples            : 
>   Differing postscript output : 
>   Missing stdout              : 
>   Differing stdout            : 
> ============================================
> 
> I can confirm the same result by manual inspection of the psc output
> files, especially example 21 which was giving us problems earlier but
> now looks exactly the same whether generated by a c/cpp code or by lua.
> I believe we can now treat plplot at the current svn revision to be
> perfectly compatible with lua 5.2 as well. Thanks a bunch for all of
> your help with the tests, building etc. and for fixing everything too.
> 
> There is an unrelated problem about the GNAT library not being detected
> by the build system, but I will report that in a separate mail. Please
> let me know if you think any more tests could help with the lua 5.2
> issue.

Dear Atri,

Many thanks for your report. It is good to know that you have
succeeded. Also, thanks for the patch to fix the strncat arguments.
I have committed this to the svn repository now. This is a common mistake 
that has slipped through the net. I'll take a quick look to see if there 
are any other occurences in the source.

Regards

Andrew