Closed xiota closed 11 months ago
It's great you made the list! I'm assuming that most of these are packages that never succeeded building after a full repository rebuild. Looking into the packages build logs sounds like a great idea, before eventually dropping them.
One of the missing packages is manim
. I noticed it is in a section that belongs to @RustemB, but that user has no activity on GitHub since the end of 2021 (corresponding with a geopolitical event in the region).
Should manim
be dropped or fixed? Since it's a broken package, there are probably no download stats to gauge interest.
binfmt-qemu-static
Is superseeded by qemu-user-static-binfmt https://archlinux.org/packages/extra/x86_64/qemu-user-static-binfmt/ if I recall well
woeusb-gui
is now woeusbgui
dmenu-height
Removed as personal build of dmenu
chrome-gnome-shell
Now is gnome-browser-connector
sneedacity
Removed from AUR https://lists.archlinux.org/archives/search?mlist=aur-requests%40lists.archlinux.org&q=sneedacity
mauikit-git
Need mauiman-git at the moment
maui-shell-git
Need other packages
One of the missing packages is
manim
. I noticed it is in a section that belongs to @RustemB, but that user has no activity on GitHub since the end of 2021 (corresponding with a geopolitical event in the region).
This has another very sad cause, our fellow co-maintainer colleague passed away during that time.
Should
manim
be dropped or fixed? Since it's a broken package, there are probably no download stats to gauge interest.
I recommend dropping it therefore.
@xiota Unfortunately RustemB passed away Dec 2021 😞…
If manim
no longer builds, then it's a candidate for removal. This might be simply due to a dependency not being in routine yet, so that should be investigated. It wouldn't be unreasonable to open a new issue for that package alone. You are correct that there are no metrics for downloads due to the time it hasn't been building. If it requires manual intervention I think it definitely should be removed.
@FabioLolix Thank you!
woeusbgui
: is the old unmaintained version, so it should stay droppedchrome-gnome-shell
: had no downloads in a month, so I think we should probably keep it dropped unless it's requested.mauikit-git
: This is still in AUR and routine, is anything wrong with it?maui-shell-git
: Same as mauikit-git
, is there a cause for this to be removed?mauikit-git: This is still in AUR and routine, is anything wrong with it? maui-shell-git: Same as mauikit-git, is there a cause for this to be removed?
Not to be removed, it was indicated that there is a build log about it (and it is failing to build), new dependencies are missing
For mauikit-git:
error: target not found: mauiman-git
But seems it can be reverted to mauiman
For mauiman-git:
error: target not found: cask-server-git
error: target not found: maui-core-git
error: target not found: mauikit-calendar-git
error: target not found: mauikit-git
error: target not found: mauiman-git
@Technetium1 Sorry about the passing of your colleague.
I took a look at trying to fix manim
. It would require adding over half a dozen new dependencies. Will remove it in my next PR.
For PRs fixing packages, should there be a separate PR for each fixed package? Or can they be grouped like removals? I'm thinking group them if they require only small changes. Otherwise there could be over a hundred PRs.
Could I get usage stats for amdgpu-fan
? I'm thinking it's a candidate for removal along with amdgpu-pro-installer
. The -installer
package looks like it has recurrent issues related to linux kernel versions. The -fan
package AUR page has a comment noting that it's old, unmaintained, and doesn't work properly.
Could I get usage stats for amdgpu-fan?
41 downloads during the last 30 days. For comparison, our 25th most download package was downloaded, 7847 times.
Could I get usage stats for amdgpu-fan?
41 downloads during the last 30 days.
Thank you.
For comparison, our 25th most download package was downloaded, 7847 times.
Now that you've mentioned it... What are the top packages and their stats?
Some packages have essentially empty logs. What's going on? For example, doomseeker
(log):
Building package "doomseeker"
:: Synchronizing package databases...
core downloading...
extra downloading...
community downloading...
multilib downloading...
chaotic-aur downloading...
:: Starting full system upgrade...
there is nothing to do
umount: /var/cache/chaotic/routines/hourly/doomseeker/machine/root/home/main-builder/pkgwork unmounted
umount: /var/cache/chaotic/routines/hourly/doomseeker/machine/root/home/main-builder/.ccache unmounted
umount: /var/cache/chaotic/routines/hourly/doomseeker/machine/root/home/main-builder/pkgsrc unmounted
umount: /var/cache/chaotic/routines/hourly/doomseeker/machine/root/var/cache/pacman/pkg unmounted
umount: /var/cache/chaotic/routines/hourly/doomseeker/machine/root/var/pkgdest unmounted
umount: /var/cache/chaotic/routines/hourly/doomseeker/machine/root (overlay) unmounted
real 0m1.846s
user 0m0.453s
sys 0m0.403s
I had this happen a few times in the past, in all cases this indicated a somehow broken PKGBUILD. It might or might not be the case here.
@xiota if it's for many things at once, it can be easier to review when they're separated. When something is initially added or removed there's a human interaction behind it. Given what's required I agree with removing manim
.
Top 50 last month:
The build log for decaf-emu-git
has this error:
==> ERROR: /home/main-builder/pkgsrc/cnl is not a clone of https://github.com/johnmcfarlane/cnl
Aborting...
The package cache needs to be cleared because the submodules were recently fixed.
Sources should be cleaned and the package built successfully.
tilix-git
is using a PKGBUILD from chaotic-aur/pkgbuild-tilix-git. Looks like it was created because of issues with the AUR package at the time. Since it's currently not building, can it be returned to the AUR version, and if needed, an interfere added?
tilix-git
is using a PKGBUILD from chaotic-aur/pkgbuild-tilix-git. Looks like it was created because of issues with the AUR package at the time. Since it's currently not building, can it be returned to the AUR version, and if needed, an interfere added?
Yes, I agree on this.
decaf-emu-git
... The package cache needs to be cleared because the submodules were recently fixed.Sources should be cleaned and the package built successfully.
I am able to build the package locally in a clean chroot, but log still shows same error today. How to clean the sources? Manual intervention required?
Actually, this should happen automatically since https://github.com/chaotic-aur/toolbox/commit/15cb300ea48ee77ad1a2430b6efd61ac0e4afb4f, though this package gets built on an HPC which works slightly differently than our "regular" builders. I thought I dropped the source cache of it yesterday :eyes: @thotypous or @PedroHLC one of you guys have time to look into it?
one of you guys have time to look into it?
Deleted it from c22 and c19. Should be fine now.
The date stamp on decaf-emu-git
log changed, but the log is still showing the same error.
==> ERROR: /home/main-builder/pkgsrc/cnl is not a clone of https://github.com/johnmcfarlane/cnl
Any idea what is going on?
My guess would be another cXX
still having the cache. I moved it to chaotic-dragon
where cache cleaning is ensured.
Some packages fail checks, like rofi-lbonn-wayland.log. Is there a way to make a PKGBUILD
skip checks by default unless --check
is passed to makepkg
?
We can omit the checks by replacing the check()
function via interfere & PKGBUILD.append
, like this: https://github.com/chaotic-aur/interfere/blob/1a4b421b7f6fdfcf07655e77531af9713565e23d/guestfs-tools/PKGBUILD.append#L1
What's going on on rofi-lbonn-wayland ?
[10/137] Compiling C object subprojects/libnkutils/nk-git-version.p/src_git-version.c.o
../rofi/subprojects/libnkutils/src/git-version.c: In function ‘_nk_git_version_run_git’:
../rofi/subprojects/libnkutils/src/git-version.c:134:5: warning: ‘g_spawn_check_exit_status’ is deprecated: Use 'g_spawn_check_wait_status' instead [-Wdeprecated-declarations]
134 | if ( ! g_spawn_check_exit_status(status, &error) )
| ^~
In file included from /usr/include/glib-2.0/glib.h:84,
IMO check failures should be investigated (almost always if it isn't an extreme hassle)
It can be useful for upstream to know they're failing checks in the first place, even if the end result is valid, and we do skip that check for now with interfere.
... replacing the
check()
function via interfere... check failures should be investigated...
Other packages have similar issues. Would prefer upstream fix or upstream PKGBUILD bypass checks.
It can be useful for upstream to know they're failing checks...
For rofi-lbonn-wayland
, the PKGBUILD maintainer hasn't acted on the check error in ~7 months. The issue has been fixed in the upstream source, but a new release hasn't been tagged. Perhaps switch to rofi-lbonn-wayland-git
?
vte3-git
builds in clean chroot on my machine, but fails with vte3-git.log:
==> Starting pkgver()...
fatal: not a git repository (or any of the parent directories): .git
==> ERROR: A failure occurred in pkgver().
unflac
also builds on my machine, but fails with with a truncated log unflac.log.
unflac
also builds on my machine, but fails with with a truncated log unflac.log.
This usually happens when something about the PKGBUILD is not okay. In this case, I found this to be the cause. Probably need to alter the function in order to pass our builders. No luck in finding the cause for the second one yet, looks like a toolbox bug to me.
Sorry I missed that... New unflac.log indicates successful build. Did you bypass check()
? I don't see an interfere.
Just found this: What do three dots "./..." mean in Go command line invocations?
qlog-git
aur has been updated, but isn't rebuilding. Log file date is 2023-05-04.
Did you bypass
check()
? I don't see an interfere.
I removed it from the PKGBUILD locally in order to build that package. Still needs an according interfere.
qlog-git
aur has been updated, but isn't rebuilding. Log file date is 2023-05-04.
Is this one on ufscar-hpc
by chance?
Is this one on ufscar-hpc by chance?
Yes, it is. I've noticed some other packages also aren't updating. I guess ufscar-hpc is having problems.
Lets see whether this solved the issue: https://github.com/chaotic-aur/packages/issues/2459#issuecomment-1542535361
@Edu4rdSHL nvidia-vaapi-driver-git
has been merged into libva-nvidia-driver-git
. PRQ#40508
@Edu4rdSHL zenstates
in CatBuilder/hourly.txt
doesn't exist. AUR has zenstates-git
and zenstates-openrc
.
@xiota ty for notifying, fixed in https://github.com/chaotic-aur/packages/commit/1696906bba55e45731f6b331a11e6fc35f008f7c
There are 200-some missing packages. I'll prepare some PRs to try to remove or fix them in stages. Since my list contains some false positives, I'll do my best to double check before opening each PR. Feel free to let me know whether any specific packages should be fixed, definitely removed, or anything else.