LLVM, LLVM-Libs, Clang and lld need a rebuild #54
Labels
No Label
blocked upstream
bug
build-failure
duplicate
enhancement
help wanted
informational
invalid
invalid/corrupt package
packaging issue
priority: high
question
support
wontfix
No Milestone
No project
No Assignees
3 Participants
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: ALHP/ALHP.GO#54
Loading…
Reference in New Issue
No description provided.
Delete Branch "%!s(<nil>)"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
LLVM, LLVM-Libs, Clang and lld pacakge need a rebuild, it cause it to crash and there a serious size difference for LLVM
That could be one of the sideffects mentioned in https://git.harting.dev/anonfunc/ALHP.GO/issues/52#issuecomment-695.
I'll requeue them and monitor if the LTO linking is OOM-killed.
Removed and requeued, lets see how it goes.
Okay it seems to have fixed some issue, but i'm in dought concerning llvm, it still have a huge difference in size, when i have some time, i will try to compile it my self and see what going on.
But llvm-libs seem to be causing some issue at compile time, maybe should we try a rebuild with lto, and if the issue seem the same on without lto ?
I agree, we need to investigate if its caused by LTO. I would blacklist it for now, until we got that figured out.
Did you try using the 'smaller' ALHP version to see if it works as intended? The size difference must come from somewhere, but it would be interesting to see if it still does its job (I think mesa uses aco for shaders by default now, so that's out).
Atm I have a package that compiling with llvm from ALPH but needed to install llvm-libs from the arch repo to make it compile, so will see if there something wrong or not, but llvm seem fine (for now).
The rust package also needs to be rebuilt I believe.
@titaniumtown
rust
seems to fail currently. Not sure if a rebuild yields new result there.@Gontier-Julien there not currently no way to blacklist a split-package. Since
llvm-libs
is part ofllvm
, blacklistingllvm
is the only option.Well I just got errors trying to compile a rust project in relation to the llvm version. idk
And also, just noticed that rust shouldn't be built with LTO:
dde13f77f3/sys-config/ltoize/files/package.cflags/lto.conf (L32)
Maybe that's the issue?
Could be connected to that weird
llvm
version that is build currently. I just blacklisted it, can you try again after downgradingllvm
(and its split packages) to arch's official version? (pacman -Suuy
)That is probably why it has failed to build, but does not explain your build failure related to
llvm
.EDIT: Interestingly enough,
llvm
is also listed there:dde13f77f3/sys-config/ltoize/files/package.cflags/lto.conf (L89)
EDIT2: Probably time to go through that list... Would need to add some code that would make it possible to disable LTO for specific packages.
The exact error I get is:
Well, your rustc is linking to llvm12, so no suprises there. Whats your rust package version? Should be
1:1.56.1-3
from [extra].It's using the rust version from this repo (
rust-1:1.56.0-1.1
) I have my pacman.conf structured like in the README.md https://git.harting.dev/anonfunc/ALHP.GO#example-pacman-confAh, rust was not removed from ALHP when it failed to build. Weird. Going to remove that shortly, then you should get the newest official version.
Got it, thank you so much!
rust
removed, should be fixed now.Yup just seen it I'm going to do the test and report back to you if it is lto or not, it just take a long time to
@Gontier-Julien I implemented a separate LTO blacklist in
de3ca80aab
.Currently on there are
rust
,llvm
andxf86-intel-video
.Awesome!
Also I've remarked that
gcc
is on the blacklist, I think it for stability reason ?gcc
behaves really weird when you try to build local packages with the x86-64-v3 version of it. I don't have the exact errors anymore, since I noticed that really early on while working on the now-deprecated python version of alhp (so over 12 months ago).Something about local header includes not matching compiled or build-in versions.
anonfunc referenced this issue2021-11-13 21:37:29 +01:00
Maybe since there is the new version, a test build couldn't hurt?
What do you think?
Probably a local build then, but yea why not. Building
gcc
local + trying to build an AUR package withgcc
as compiler should be enough to test that.@anonfunc do you think it would be a good idea to create a Discord server for this project (if there isn't one already)? So things like this could be discussed there, was just an idea.
Could be a great idea @titaniumtown ^^
So a little report on the test with just llvm-libs from arch.
The package with llvm build successfully and work as expected no bug, my test build of llvm never finished/is taking to long
So for me, the size difference didn't cause issues, but llvm-libs did, so could be that the lto does someting that reduce a lot of code and/or doc
llvm take agggeeee to build my gosh
@titaniumtown maybe could you try to build it on your end?
Btw @anonfunc the blacklisted LTO packages is being build without LTO right? Because if that so LLVM wasn't rebuild without LTO
It'll be build if I requeue it or a new version appears. I may requeue it later.
Okay no problem ^^
Closing this on the same basis as #57, please reopen if new package versions still do not work correctly.