Blacklisted packages #16

Open
opened 2021-07-09 20:02:55 +02:00 by anonfunc · 28 comments
Owner

Here are some that are going to stay with specific reasons:

  • tensorflow, pytorch: takes ages to build and mountains of disk space that I simple do not have to spare on the buildserver
  • tensorflow-cuda: see tensorflow
  • gcc: Would blow up your local build process for e.g. AUR packages
  • electron*: Each version takes a lot of buildtime, and there are 8 versions of electron in the repo (for each march!) at the time of writing this.

Current list:

Packages build without LTO

Here are some that are going to stay with specific reasons: * tensorflow, pytorch: takes ages to build and mountains of disk space that I simple do not have to spare on the buildserver * tensorflow-cuda: see tensorflow * gcc: Would blow up your local build process for e.g. AUR packages * electron*: Each version takes a _lot_ of buildtime, and there are 8 versions of electron in the repo (for each march!) at the time of writing this. Current list: - tensorflow - tensorflow-cuda - python-pytorch - python-pytorch-cuda - python-pytorch-opt - python-pytorch-opt-cuda - electron* - gcc - dhclient (#20) - [ctags](https://somegit.dev/anonfunc/ALHP.GO/issues/16#issuecomment-208) - bluez (#31) - bandwhich (#45) - mold (https://somegit.dev/ALHP/ALHP.GO/issues/136#issuecomment-1872) - julia (https://somegit.dev/ALHP/ALHP.GO/issues/16#issuecomment-1935) - graphviz (https://somegit.dev/ALHP/ALHP.GO/issues/229#issuecomment-3132) [Packages build without LTO](https://github.com/InBetweenNames/gentooLTO/blob/435a9d968854fef21015796a5f464243dc4caa03/sys-config/ltoize/files/package.cflags/lto.conf)
anonfunc added the
enhancement
label 2021-07-09 20:03:03 +02:00
anonfunc added this to the beta milestone 2021-07-09 20:03:12 +02:00
anonfunc changed title from Cut down on blacklisted packages to Blacklisted packages 2021-07-09 20:12:01 +02:00
anonfunc added a new dependency 2021-07-09 22:47:04 +02:00
anonfunc removed a dependency 2021-07-09 22:47:17 +02:00
Author
Owner

I just removed

  • texi2html
  • gradle
  • re2
  • basket
  • w3m
  • ctags
  • libedit
  • jre-openjdk-headless
  • jre-openjdk
  • qca
  • skia-sharp58

and added

  • dhclient (#20)
I just removed * texi2html * gradle * re2 * basket * w3m * ctags * libedit * jre-openjdk-headless * jre-openjdk * qca * skia-sharp58 and added * dhclient (#20)
anonfunc added
question
and removed
enhancement
labels 2021-07-10 00:05:52 +02:00
anonfunc removed this from the beta milestone 2021-07-10 01:33:31 +02:00
Author
Owner

Added back skia-sharp58, which is some sort of Schroedinger package: exists and not exists simultaneously.

svn2git vs archweb

Added back ctags, some weird version ctags-1:r20210106+g08b1c490c-1 was being build, over and over again. Needs more investigating. Maybe the combination of +git and epoch breaks something.

Added back `skia-sharp58`, which is some sort of *Schroedinger* package: exists and not exists simultaneously. [svn2git](https://github.com/archlinux/svntogit-community/tree/master/skia-sharp58/repos/community-x86_64) vs [archweb](https://archlinux.org/packages/?q=skia-sharp58) Added back `ctags`, some weird version `ctags-1:r20210106+g08b1c490c-1` was being build, over and over again. Needs more investigating. Maybe the combination of +git and epoch breaks something.
Author
Owner

Removed pacman, see #41 for more.

Removed `pacman`, see #41 for more.
Author
Owner

Added chezmoi, see #43 for details.

Added `chezmoi`, see #43 for details.
Author
Owner

Added bandwhich, see #45.

Added `bandwhich`, see #45.
Author
Owner

Added rust, llvm and xf86-video-intel to build whithout LTO.

See https://git.harting.dev/anonfunc/ALHP.GO/issues/54#issuecomment-727 for more.

Added `rust`, `llvm` and `xf86-video-intel` to build whithout LTO. See https://git.harting.dev/anonfunc/ALHP.GO/issues/54#issuecomment-727 for more.
Author
Owner

Removed skia-sharp58 and chezmoi, since ALHP can now handle orphaned PKGBUILD files.

Removed `skia-sharp58` and `chezmoi`, since ALHP can now handle orphaned PKGBUILD files.
anonfunc added
informational
and removed
question
labels 2021-11-22 23:56:22 +01:00

One question, is php-fpm blacklisted from the repo for any cause in particular?

I'm asking because a lot of php related packages are available but not that one.

Edit: Same for php-fpm7

One question, is php-fpm blacklisted from the repo for any cause in particular? I'm asking because a lot of php related packages are available but not that one. Edit: Same for php-fpm7
Author
Owner

php-fpm should not be blacklisted, and in current version ALHP it can't be, because it's part of php as a split-package, and these can't be blacklisted.

As far as I can see, php-fpm is available in both x86-64-v2 and x86-64-v3:

Repository      : extra-x86-64-v3
Name            : php-fpm
Version         : 8.0.14-1.1
Description     : FastCGI Process Manager for PHP
Architecture    : x86_64
URL             : https://www.php.net/
Licenses        : PHP
Groups          : None
Provides        : None
Depends On      : php  systemd
Optional Deps   : None
Conflicts With  : None
Replaces        : None
Download Size   : 2.72 MiB
Installed Size  : 16.07 MiB
Packager        : ALHP x86-64-v3 <alhp@harting.dev>
Build Date      : Sun 19 Dec 2021 19:04:37 CET
Validated By    : MD5 Sum  SHA-256 Sum  Signature
`php-fpm` should not be blacklisted, and in current version ALHP it can't be, because it's part of `php` as a split-package, and these can't be blacklisted. As far as I can see, `php-fpm` is available in both `x86-64-v2` and `x86-64-v3`: ``` Repository : extra-x86-64-v3 Name : php-fpm Version : 8.0.14-1.1 Description : FastCGI Process Manager for PHP Architecture : x86_64 URL : https://www.php.net/ Licenses : PHP Groups : None Provides : None Depends On : php systemd Optional Deps : None Conflicts With : None Replaces : None Download Size : 2.72 MiB Installed Size : 16.07 MiB Packager : ALHP x86-64-v3 <alhp@harting.dev> Build Date : Sun 19 Dec 2021 19:04:37 CET Validated By : MD5 Sum SHA-256 Sum Signature ```

Oh my god, I'm so dumb.

I was checking it with the packages.html and it's not there, but it is effectively in the repo.

I'm sorry, you are correct.

Oh my god, I'm so dumb. I was checking it with the packages.html and it's not there, but it is effectively in the repo. I'm sorry, you are correct.
Author
Owner

Added julia, since its checks take an enormous amount of ram. This may be a candidate for a no-checks-list in the future.

Added `julia`, since its checks take an enormous amount of ram. This may be a candidate for a *no-checks-list* in the future.

@anonfunc what do you mean compiling gcc for x86 64-v3 would blow up the build process?

@anonfunc what do you mean compiling gcc for x86 64-v3 would blow up the build process?
Author
Owner

@diegor Meaning it broke building locally (AUR packages for example) since gcc gets version mismatches in its plugins.

@diegor Meaning it broke building locally (AUR packages for example) since gcc gets version mismatches in its plugins.

@anonfunc What do you mean by plugins? Compiler front ends?

@anonfunc What do you mean by plugins? Compiler front ends?
Author
Owner

@GunpowderGuy Don't think so. This was many years ago, but the different versioning somehow broke local compiling. If you want to test it again, be my guest, but I doubt this is has changed without us doing anything about that.

@GunpowderGuy Don't think so. This was many years ago, but the different versioning somehow broke local compiling. If you want to test it again, be my guest, but I doubt this is has changed without us doing anything about that.

@anonfunc Do you have sources that can help me better understand this problem ? Someone else ought to have had it.

Improving the performance of gcc is already desirable but particularly in arch due to AUR

@anonfunc Do you have sources that can help me better understand this problem ? Someone else ought to have had it. Improving the performance of gcc is already desirable but particularly in arch due to AUR
Author
Owner

You can have a look at ALHP's makepkg.confs here. Use the same settings, pull gcc with asp checkout gcc and build it with makepkg or extra-x86_64-build. After you have installed it, try compiling some gcc based aur or repo packages and see what happens.

You can have a look at ALHP's makepkg.confs [here](https://alhp.dev/makepkg/). Use the same settings, pull `gcc` with `asp checkout gcc` and build it with `makepkg` or `extra-x86_64-build`. After you have installed it, try compiling some gcc based aur or repo packages and see what happens.

Why OpenCV and related packages are blacklisted?

Why OpenCV and related packages are blacklisted?
Author
Owner

Good question, I don't quite remember. I'll unlist it, maybe I can find out why I had it blacklisted (or maybe it does build fine now).

EDIT: Seems to fail: https://github.com/opencv/opencv/issues/23893

Good question, I don't quite remember. I'll unlist it, maybe I can find out why I had it blacklisted (or maybe it does build fine now). *EDIT*: [Seems to fail](https://alhp.dev/logs/x86-64-v3/opencv.log): https://github.com/opencv/opencv/issues/23893
anonfunc pinned this 2023-07-17 17:26:31 +02:00

It seems that networkmanager is failed to build and libnm package is missing as a result.

It seems that networkmanager is [failed to build](https://somegit.dev/ALHP/ALHP.GO/issues/201) and [libnm](https://archlinux.org/packages/extra/x86_64/libnm/) package is missing as a result.
Author
Owner

@entrider Should be provided from the official repos as usual.

@entrider Should be provided from the official repos as usual.

Probably xxhash should be blacklisted just for v3: https://alhp.dev/logs/x86-64-v3/xxhash.log
Builds fine with x86-64-v2 tho.

Also gstreamer build failed with x86-64-v3, but not with v2.

Probably xxhash should be blacklisted just for v3: https://alhp.dev/logs/x86-64-v3/xxhash.log Builds fine with x86-64-v2 tho. Also `gstreamer` [build failed](https://alhp.dev/logs/x86-64-v3/gstreamer.log) with x86-64-v3, but not with v2.

Are telegram-desktop, qt5-webengine and nodejs blacklisted? Why?

Are `telegram-desktop`, `qt5-webengine` and `nodejs` blacklisted? Why?
Author
Owner

Probably xxhash should be blacklisted just for v3: https://alhp.dev/logs/x86-64-v3/xxhash.log
Builds fine with x86-64-v2 tho.

We can just leave it as failed. This way new upstream versions get build and we can see if anything is improving.

Also gstreamer build failed with x86-64-v3, but not with v2.

See #201. Also, this issue is really not the pace to discuss failing packages. If you open any upstream issues for failing packages, feel free to open a corresponding issue here so we can track its progress.

Are telegram-desktop, qt5-webengine and nodejs blacklisted? Why?

They are. Probably because they took too long to build. With a more powerful buildserver we could add them back, but that is not an option right now.

> Probably xxhash should be blacklisted just for v3: https://alhp.dev/logs/x86-64-v3/xxhash.log > Builds fine with x86-64-v2 tho. We can just leave it as failed. This way new upstream versions get build and we can see if anything is improving. > Also `gstreamer` [build failed](https://alhp.dev/logs/x86-64-v3/gstreamer.log) with x86-64-v3, but not with v2. See #201. Also, this issue is really not the pace to discuss failing packages. If you open any upstream issues for failing packages, feel free to open a corresponding issue here so we can track its progress. > Are `telegram-desktop`, `qt5-webengine` and `nodejs` blacklisted? Why? They are. Probably because they took too long to build. With a more powerful buildserver we could add them back, but that is not an option right now.

Can mold be reconsidered? currently make test passes for me in 35s with x86-64-v3 enabled (locally at least)

Can `mold` be reconsidered? currently `make test` passes for me in 35s with x86-64-v3 enabled (locally at least)

They are. Probably because they took too long to build. With a more powerful buildserver we could add them back, but that is not an option right now.

Is there any way we can help to make this happen? either host a powerful server, buy you a threadripper, or donate X amount of dollars to make this economically viable

> They are. Probably because they took too long to build. With a more powerful buildserver we could add them back, but that is not an option right now. Is there any way we can help to make this happen? either host a powerful server, buy you a threadripper, or donate X amount of dollars to make this economically viable
Author
Owner

They are. Probably because they took too long to build. With a more powerful buildserver we could add them back, but that is not an option right now.

Is there any way we can help to make this happen? either host a powerful server, buy you a threadripper, or donate X amount of dollars to make this economically viable

I actually added back nodejs on the newer server already. Maybe that can be done for webkit as well, I'll have a look. As for the buildserver hardware: It's currently powered by a Ryzen 7600, so of course there is more performance to be had.
Cheapest Threadripper with AVX512 you can get here is about $1800 + $800 mainboard. May be a little overkill (7960X, 24C/48T).

That's obviously quite a bit of money to spend on something like this, so I would not ask that of anybody. Just giving you a figure here.

Can mold be reconsidered? currently make test passes for me in 35s with x86-64-v3 enabled (locally at least)

I'll test mold again once I'm back home, currently traveling.

> > They are. Probably because they took too long to build. With a more powerful buildserver we could add them back, but that is not an option right now. > > Is there any way we can help to make this happen? either host a powerful server, buy you a threadripper, or donate X amount of dollars to make this economically viable I actually added back nodejs on the newer server already. Maybe that can be done for webkit as well, I'll have a look. As for the buildserver hardware: It's currently powered by a Ryzen 7600, so of course there is more performance to be had. Cheapest Threadripper with AVX512 you can get here is about $1800 + $800 mainboard. May be a little overkill (7960X, 24C/48T). That's obviously quite a bit of money to spend on something like this, so I would not ask that of anybody. Just giving you a figure here. > Can `mold` be reconsidered? currently `make test` passes for me in 35s with x86-64-v3 enabled (locally at least) I'll test mold again once I'm back home, currently traveling.
Author
Owner

@AvianaCruz Can you open a new issue for this please? This issue is meant to track the list of packages we exclude from building.

@AvianaCruz Can you open a new issue for this please? This issue is meant to track the list of packages we exclude from building.
Sign in to join this conversation.
No description provided.