Free Software Activities in August 2018
Hello all! After reading some of the nice activity logs published by members of the Debian Project at https://planet.debian.org, I've decided to publish my own version. This will mostly be my work on FreeCAD and Debian, but will also include other points of note from my many interests. I'll also include relevant activities for my supporters on Patreon
As covered in my previous post, most of the first half of August was spent finishing my Google Summer of Code project for FreeCAD, which largely turned into "packaging for the Debian Science Team" since FreeCAD's dependency chain and options for integration are so vast. However, the nature of the work is highly parallel. For example, when packaging, most issues occur either almost immediately into the build, or after the build has finished. This means that after an initial bit of packaging work that allows a build to actually proceed, one then has to launch the build and wait anywhere from a few minutes to a few hours to see the next issue arise. So, as a result, I worked on many packages in parallel. A negative consequence of this was that I somewhat overloaded my plate in terms of things I can handle, so in actuality this meant I would work on a handful of packages on several machines at once, and then once those issues were largely resolved I would move onto another handful, rather than really juggling them all at once.
After finishing GSoC, I first turned my attention to two of the packages I had worked on in parallel, python-fluids and python-ulmo.
python-fluids
I made a new upload for this package to correct several issues. Here's an excerpt from the changelog:
* [1215102] Make python-fluids-doc Multi-Arch: foreign * [eb610fd] Correct Vcs-* fields * [62570cd] Updated to Standards-Version 4.2.0, no change required * [29e96a7] Correct debian/watch * [3e90d05] Add missing scipy dependency * [cd776a2] Add lintian override for package-contains-documentation-outside-usr-share-doc
The package still fails some reproducibility tests on i386 and could use a standards version bump, but other than that, this package is almost completely clean now! It's quite satisfying cleaning up these packages.
python-ulmo
I made an upload for this package as well:
* [7150daa] Correct Vcs-* fields * [027b50b] Mark python-ulmo-doc as Multi-Arch: foreign * [d72b6f6] Add patch to fix FutureWarning
The package was failing its autopkgtests because Python was emitting a FutureWarning onto stderr. This package is actually fully reproducible, and only needs a standards version bump to be fully clean.
gmsh
On August 22nd, Gmsh 4.0.0 was released! The Java API was removed and a Julia one added. This is a nice release because in Debian bug #904946 it was found that the Gmsh package, after being updated to be built against OCCT instead of OCE, actually no longer worked when used in FreeCAD, giving the error:
Error : Gmsh requires OpenCASCADE to import shape
My hope was that by updating to the new Gmsh version, this problem would resolve itself. However, I ran into a few problems and had to postpone finishing up.
netgen
On August 19th, Debian bug #906619 was reported against Netgen, which was an occasion to look at the state of the package. I noticed that upstream had released version 6.2.1807 (versus the current .1804) but it was not tagged in the Netgen repository, only as a submodule of the Ngsolve repository.
On August 31st, I packaged the new version and resolved the bug in science-team/netgen PR #9.
salome-smesh
Although technically not August, on September 3rd I pushed the nearly finished package for a standalone salome-smesh on Salsa which is fully functional except for an issue with the CMake config file. I submitted a pull request to support multi-arch for the package but forgot to update the path in the CMake config file, so the variables therein have to be manually set, or the file edited.
cantera
On August 24th, Cantera 2.4.0 was released:
Cantera is an open-source suite of tools for problems involving chemical kinetics, thermodynamics, and transport processes.
Previously there had been an issue with the build, where the packages would be made successfully, but attempting to import cantera in Python would error out, presumably due to the way an included copy of libfmt was handled during the build. However, the issue is no longer present, so I now have a working Cantera 2.4.0 package! The only shortcoming is that I don't have working docs build yet, so this package didn't make it in time for August.
openfoam
Previously I had submitted a pull request for OpenFOAM 6 to the Debian Science Team, but spontaneously the package began to fail, presumably due to some change in a dependency. I spent some time troubleshooting this issue but haven't found a resolution yet. Here's a snippet of the build failure:
Allwmake src/Pstream
wmake dummy
wmakeLnIncludeAll: running wmakeLnInclude on dependent libraries:
wmakeLnInclude: linking include files to /build/openfoam-6.20180805+dfsg1/src/OpenFOAM/lnInclude
wmakeLnInclude: linking include files to /build/openfoam-6.20180805+dfsg1/src/OSspecific/POSIX/lnInclude
make[2]: Entering directory '/build/openfoam-6.20180805+dfsg1/src/Pstream/dummy'
wmakeLnInclude: linking include files to ./lnInclude
make[2]: Leaving directory '/build/openfoam-6.20180805+dfsg1/src/Pstream/dummy'
make[2]: Entering directory '/build/openfoam-6.20180805+dfsg1/src/Pstream/dummy'
Making dependency list for source file UOPwrite.C
realloc(): invalid next size
make[2]: *** Deleting file '/build/openfoam-6.20180805+dfsg1/platforms/linux64GccDPInt32Opt/src/Pstream/dummy/UOPwrite.C.dep'
Making dependency list for source file UIPread.C
realloc(): invalid next size
make[2]: *** Deleting file '/build/openfoam-6.20180805+dfsg1/platforms/linux64GccDPInt32Opt/src/Pstream/dummy/UIPread.C.dep'
Making dependency list for source file UPstream.C
realloc(): invalid next size
make[2]: *** Deleting file '/build/openfoam-6.20180805+dfsg1/platforms/linux64GccDPInt32Opt/src/Pstream/dummy/UPstream.C.dep'
I'll have to continue investigating this one.
pyside2
I reported Debian bug #906020 about missing Python 3 CMake config files, which was fixed Aug 28th. I began investigating how this would Pivy builds, but got sidetracked by a problem with the med-fichier package.
freecad
I ran into some build issues trying to update the package now that my Debian Unstable workstation is running Qt 5.11:
error: invalid use of incomplete type ‘class QAction’
QAction* action = new QAction(tr("Remove"), this);
There are several instances of this throughout the codebase that can basically be fixed with by adding some #include s.
projectchrono
There was a bit of discussion about utilization of Project Chrono, an open source multi-physics simulation engine, in FreeCAD. One problem with pursuing this is that it isn't packaged at all for Debian, which is rather surprising! It seems like high-quality software. I began looking at packaging it and it actually doesn't seem that bad. One of the main issues is actually the naming: it's really long! There's also a bit of worry about naming collisions I'll have to look into. As of now, I have a working package except for the Python bindings. Enabling them results in this failure in the build process:
[ 53%] Building CXX object src/demos/irrlicht/CMakeFiles/demo_IRR_soilbin.dir/demo_IRR_soilbin.cpp.o
cd /build/project-chrono-3.0.0/obj-x86_64-linux-gnu/src/demos/irrlicht && /usr/lib/ccache/c++ -DBP_USE_FIXEDPOINT_INT_32 -I/build/project-chrono-3.0.0/src -I/build/project-chrono-3.0.0/obj-x86_64-linux-gnu -I/build/project-chrono-3.0.0/src/chrono -I/build/project-chrono-3.0.0/src/chrono/collision/bullet -I/build/project-chrono-3.0.0/src/chrono/collision/gimpact -I/build/project-chrono-3.0.0/src/chrono/collision/convexdecomposition/HACD -I/usr/include/irrlicht -g -O2 -fdebug-prefix-map=/build/project-chrono-3.0.0=. -fstack-protector-strong -Wformat -Werror=format-security -Wdate-time -D_FORTIFY_SOURCE=2 -pthread -fopenmp -march=native -msse4.2 -mfpmath=sse -march=native -mavx2 -march=native -mfma -std=c++14 -pthread -fopenmp -march=native -msse4.2 -mfpmath=sse -march=native -mavx2 -march=native -mfma -std=c++14 -pthread -fopenmp -march=native -msse4.2 -mfpmath=sse -march=native -mavx2 -march=native -mfma -o CMakeFiles/demo_IRR_soilbin.dir/demo_IRR_soilbin.cpp.o -c /build/project-chrono-3.0.0/src/demos/irrlicht/demo_IRR_soilbin.cpp
/build/project-chrono-3.0.0/obj-x86_64-linux-gnu/swig/ChModuleCorePYTHON_wrap.cxx: In member function 'virtual void chrono::ChLogPython::Output(const char*, size_t)':
/build/project-chrono-3.0.0/obj-x86_64-linux-gnu/swig/ChModuleCorePYTHON_wrap.cxx:7610:23: error: format not a string literal and no format arguments [-Werror=format-security]
PySys_WriteStdout(buffer);
^~~~~~
opencamlib
There was a problem with this package's doc depends, which I fixed at the end of August. However, it's still awaiting sponsorship to be uploaded into the Debian archives.
med-fichier
The med-fichier package in Debian currently is built with autotools but somewhat supports CMake builds, which would be nice to switch to as it would provide CMake configuration files in libmedc-dev. I worked on updating the package, but currently building docs is broken, and for some reason, the Fortran library is being statically linked.
Planet FreeCAD
The planet-venus package in Debian is available for running your own RSS feed aggregation service. I wanted to set up a planet.freecad.io to provide exactly that. However, the current version of the package is broken, and it only works in Debian 7, a.k.a. old-old-stable.
I spent some time researching the state of the package, but it's fairly rough, and most importantly, upstream is far from active. I'd like to run the RSS service but fixing the package seems like a tall order.
freecad.io
We previously had two servers available to run some of the web services for the FreeCAD project, but sadly they rebooted but never came back online.
I'm now paying for a VPS in Frankfurt using the money from my Patreon. Unfortunately, it's a bit resource-strained from the FreeIPA server hogging things.
Comments
Comments powered by Disqus