| Summary: | dev-util/intel-graphics-compiler-1.0.17537.1: syntax error in VERSION script | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Nowa Ammerlaan <nowa> |
| Component: | Current packages | Assignee: | Jan Henke <gentoo> |
| Status: | CONFIRMED --- | ||
| Severity: | normal | CC: | conikost, eschwartz, gasc, proxy-maint, toralf |
| Priority: | Normal | ||
| Version: | unspecified | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
| Bug Depends on: | |||
| Bug Blocks: | 618550 | ||
| Attachments: | build log | ||
|
Description
Nowa Ammerlaan
2024-08-26 10:50:53 UTC
Created attachment 901261 [details]
build log
Got the same issue and solved it by disabling lto Version dropped from tree, so closing. Why would a bug be fixed just because the original version it was detected in has been removed from the tree? (In reply to Eli Schwartz from comment #4) > Why would a bug be fixed just because the original version it was detected > in has been removed from the tree? Can you reproduce? I cannot with current version. (In reply to Conrad Kostecki from comment #5) > Can you reproduce? I cannot with current version. Were you able to reproduce it with the old version? The upstream code with the error is still there... they generate an ld version script by objdumping .o files even though they already use visibility macros and -fvisibility=hidden, and their commands do not work when the object files have been compiled with -flto. I have not tried reproducing it because it wants me to build all of llvm 15 and it's not available on the binhost, and compiling llvm by hand is a bit more pain than I want to engage in. But I inspected the code manually and I cannot see a reason why it would have been fixed. Note that you will only be able to reproduce it if you compile with CFLAGS / CXXFLAGS containing "-flto", so by default you wouldn't normally see the issue if you don't normally compile with LTO. ... Hmm, maybe I should add older llvms explicitly to the binhost configs. (In reply to Eli Schwartz from comment #6) > (In reply to Conrad Kostecki from comment #5) > > Can you reproduce? I cannot with current version. > > Were you able to reproduce it with the old version? No. > Note that you will only be able to reproduce it if you compile with CFLAGS / > CXXFLAGS containing "-flto", so by default you wouldn't normally see the > issue if you don't normally compile with LTO. Well, if that's the cause, we just should filter lto. We do that even for intel-compute-runtime. I don't see upstream fixing this in future. They don't even finished LLVM16 support by now.. We are stuck at LLVM15. (In reply to Conrad Kostecki from comment #7) > (In reply to Eli Schwartz from comment #6) > > Were you able to reproduce it with the old version? > > No. So what did you think changed with the new version? :P (In reply to Conrad Kostecki from comment #7) > Well, if that's the cause, we just should filter lto. We do that even for > intel-compute-runtime. I don't see upstream fixing this in future. They > don't even finished LLVM16 support by now.. We are stuck at LLVM15. Surely we should also do our part by reporting the issue, whether we expect it to be solved or not? After that we should of course filter-lto since fixing that kind of deep issue is out of scope for a downstream patch. But again, that version script looks totally wrong regardless of whether it fails to compile. :) I think we should report that it's a problem regardless of anything else. *** Bug 940544 has been marked as a duplicate of this bug. *** |