Requesting a new release of windows, which is basically just updates to allow it to run on octave 5.0
commit: https://sourceforge.net/p/octave/windows/ci/92d1ed50bfbdb6cd255391e1076daad2906122ee/tree/
742c949a594e61d7c87c8ebe2e1320d7 release/windows-1.3.1.tar.gz
5cf771483efd1c3c34088e5b0591f150 release/windows-html.tar.gz
Updated to fix crash when doing a clear all on objects created from oct files.
https://sourceforge.net/p/octave/windows/ci/376557f362e4cc808f641bf064e76934747f33cc/tree/
3f52951ea0681d5da8acc6519f4aaa5a0953362a release/windows-1.3.1.tar.gz
a0bf743725beb6160a896ec60f9d22aad3da9a23 release/windows-html.tar.gz
those were sha1sum values - md5sums
8f9f82c209e69546eff954a8ace0d6de windows-1.3.1.tar.gz
628716f496bde7e6b4b77b74d4940051 windows-html.tar.gz
For reference, what I tested before upload the package here
windows 1.3.1
[x] release candidate installs on latest stable Octave release
[x] no compiler errors or warnings
[x] ran tests using runtests (fullfile(pkg('prefix'), pkgname)
(in windows, after pkg load)
[ ] ran doctest on all functions (optional)
[x] Above steps were run on Octave versions:
(in windows)
3.6.4
3.8.2
4.0.3
4.2.2
4.4.0
5.1.0
* 5.1.1 (dev)
[x] mxe cross build compiles
[ ] tested with minimum Octave version list in DESCRIPTION
3.6.0
[x] reasonable dependencies listed in DESCRIPTION
[x] ran generate_package_html
[x] no makeinfo errors and warnings during HTML build
even though it requires windows for the functions to work, the package will compile and allow generation of documentation in linux - tests will obviously fail in linux
Is there someone who can doublecheck the release on Windows? If not, I'd tend to say we rely on the tests performed by lostbard and release the package after only some formal checks.
I agree, it would be nice to have at least one independent reviewer with access to Windows go over this release candidate, @lostbard maybe you can ping Philip or Markus or other devs who regularly do Octave Windows work to go over the review checklist?
I posted a request on the maintainers list
Sure I'll do a (limited) review, mostly functionality:
== Repository ==
[x] maintainer has specified a corresponding revision commit/changeset
[x] checkout code from Sourceforge, commit/changeset is present
(inspected configure.ac manually)
== Compile and Install ==
[x] release candidate installs on latest stable Octave release
==> message "including conditional functions"
[x] no compiler errors or warnings
[ ] ran tests using
runtests /path/to/pkg
===> doesn't work on my box, hangs forever
[ ] ran all tests, including those in src (how??)
[ ] ran doctest on all functions (optional)
[ ] ran generate_package_html (if Makefile present try
make html
)[ ] no makeinfo errors and warnings during HTML build
[ ] unpacked and spot-checked the generated HTML documentation
== Interaction with pkg ==
[x]
pkg load foo
runs with errors or warnings[x]
pkg unload foo
runs with errors or warnings[x]
pkg uninstall foo
runs with errors or warningsOnly if you first do (for any OF package with binary modules):
pkg unload foo ## to unload the binary modules
clear -f ## to really remove them from memory and unlock the files
pkg uninstall foo ## then it works w/o errors
[ ] Above steps were run on Octave and OS versions:
* 5.1.0
* 5.1.1 (from March 31, 2019)
* 6.0.0 (ditto)
== Package files in release candidate tarball ==
[ ] tested with minimum Octave version list in DESCRIPTION
[x] reasonable dependencies listed in DESCRIPTION
[x] NEWS file makes sense, version and date match
[x] All functions are listed in INDEX
AFAICS
[ ] check licenses (
licensecheck -r
"plus some manual checks").[x] package files are readable/executable by all users (reasonably current maintainer Makefile should be doing this).
[ ] version number in src/configure.ac (if present) matches DESCRIPTION and tarball name.
[ ] any version numbers within the help or function body (e.g., banners) matches above.
[ ] contains no hidden dot files, junk backup files, results of configure runs, etc (should be taken care of by maintainer Makefile).
All io package functions that need the windows pkg also work.
Ah, runtest doesn't seem to like the short filename thing on windows, but:
runtests C:\Programs\Octave\Octave-6.0.0_20190331D\mingw64\share\octave\packages\windows-1.3.1
Processing files in C:\Programs\Octave\Octave-6.0.0_20190331D\mingw64\share\octave\packages\windows-1.3.1:
iscom.m ..................................................... PASS 2/2
@Philip: could you also run the tests in the subdirectory? I mean this:
runtests C:\Programs\Octave\Octave-6.0.0_20190331D\mingw64\share\octave\packages\windows-1.3.1\@octave_com_object
Performing the formal tests, I removed a left-over src/autom4te.cache directory (but this could also be removed from the root level Makefile).
But I think src/grab.cc should be explicitly put under the GPL (v3 or later). In the same run, we could, if only for consistency, change the license in those files still under GPL (v2 or later) to GPL (v3 or later). These files are:
examples/mat2xls.m: GPL
inst/mat2xls.m: GPL
src/image.m.in: GPL
I'll update the licenses
Added licence updates and changes to the maintainer makefile for making reproducable checksums, and fixed removal of autom4te.cache
https://sourceforge.net/p/octave/windows/ci/28a0921e350a6c6e73ecd7b4eff1bb086edabfcb/tree/
d23478cf859336f75a32b8cd9ee8f468 release/windows-1.3.1.tar.gz
c43ea748c3291f4219acc2be28ba432c release/windows-html.tar.gz
Olaf you mean this?
@Philip: Yes, thanks!
@lostbard: Ok, starting to push the release online now...
...ready pushing the release. (Can be announced at the mailing lists.)
Thanks!