Gentoo Archives: gentoo-soc

From: listout <brahmajit.xyz@×××××.com>
To: gentoo-soc <gentoo-soc@l.g.o>
Cc: sam <sam@g.o>, dilfridge <dilfridge@g.o>
Subject: [gentoo-soc] Week 4 Report for Musl support expansion to support GNOME desktop
Date: Sun, 10 Jul 2022 16:06:01
Message-Id: 20220710160550.gyiqqysiuxtoyh22@gmail.com
1 Hi folks, hope everyone is doing well.
2
3 Week 4 has ended, hence I'm here with my 4th weekly report. Nothing
4 notable to report though. I have been using my patched GNOME and MATE
5 desktop environments on musl profile as my daily driver for the past
6 week, and haven't noticed bugs or anomalies.
7
8 With significant parts of the GNOME and MATE patched I started working
9 on miscellaneous packages that were failing on musl. Mainly those
10 packages needed a patch to not include either the execinfo.h or error.h
11 headers and the function that comes with those headers. Some of notable
12 packages facing [now patched] are debugedit [1], google-perftools [2],
13 netcdf [3], geant [4], and last but not the least hpx [5]. I'm sure more
14 such packages are to come up and I'll keep on fixing them. But I'm
15 however not too happy with the way I'm fixing these issues, i.e.
16 checking if `__GLIBC__` is defined, another reason being unhappy with
17 this kind of patches are can't be upstreamed. There is better way of
18 going around these issue, one of them being using either configure.ac or
19 CMakeLists.txt to determine whether execinfo.h is available and set a
20 include header guard, which in turn can be used to either include or not
21 include those files.
22
23 I've talked with my mentor regarding if we can include libexecinfo for
24 musl as Void and Alpine Linus seems to providing it as a result these
25 patches are not required there. We do plan on taking further look at
26 this, just not for now, once we have completed a significant portion of
27 our proposal then we can focus on this particular issue.
28
29 However not everything is needs to be sad in this report, I've two good
30 news to share. I now have learned how to version bump a package. The
31 google-perftools package recently fixed an issue that would fix building
32 on musl. So instead of patching the current version, I decided to
33 version bump google-perftools ebuild to 2.10 from 2.9.1, and one of my
34 patches [6] landed upstream and merged. So going on forward we won't be
35 needed the patches for hpx [5]. So, yay! for that.
36
37 However, there are still things needed to be done from my end. Midterm
38 evaluations are coming up and I've got my fingers crossed about that.
39 The plan of *keeping track of bugs that might cause run time issues for
40 other libc* needs to be worked on, I've yet to create detailed
41 documentation of the bugs/issues that I came across, this was part of my
42 project proposal. I'm still trying to set up GRS, and I agree with
43 *blueness*, it's a lot of work. The plan on killing the ::musl needs to
44 be worked too.
45
46 So for the next few weeks, I'll work with my mentors to get at least one
47 of those aforementioned parts of my proposal done.
48
49 That will be it for this week's report. Signing off.
50
51 [1]: https://github.com/gentoo/gentoo/pull/26223
52 [2]: https://github.com/gentoo/gentoo/pull/26244
53 [3]: https://github.com/gentoo/gentoo/pull/26284
54 [4]: https://github.com/gentoo/gentoo/pull/26313
55 [5]: https://github.com/gentoo/gentoo/pull/26281
56 [6]: https://github.com/STEllAR-GROUP/hpx/pull/5947
57 --
58 Regards,
59 listout