[Bug server/26839] Systemtap build failures with clang

[tbaeder at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=26839

--- Comment #10 from Timm Bäder <tbaeder at redhat dot com> ---
(In reply to Aaron Merey from comment #9)
> (In reply to Timm Bäder from comment #8)
> > Hey Frank do you have any update on this? We've been carrying the patches
> > around for a while now and they need maintenance and start to collide with
> > upstream changes, e.g.
> > https://sourceware.org/git/?p=systemtap.git;a=commit;
> > h=439fb4cc4c08166dccb85ba202a5762c4c46ba42
> > 
> > I think only the probe::printsig() problems need clarification, the other
> > ones are straight forward.
> 
> Hi Timm. When I merged that commit I didn't realise you had a similar patch
> pending. Regarding the printsig patch, I would suggest dropping the 'bool
> nest = true' definitions and instead call printsig_nested unconditionally.

No problem, in the end it's one patch less we carry around. Dropping the nest
bool is fine with me of course, I have a patch doing that. I can't seem to find
where to send patches however, the mailing list seems to have remarkably few of
them and the HACKING document doesn't seem to mention? Where does systemtap
development happen?

> Otherwise the patch set looks ok. With it, gcc still compiles systemtap
> without error and we are closer to being able to compile systemtap with
> clang. I still get an errors when compiling with clang however. ex.
> `unsupported argument 'auto' to option 'flto='` when compiling
> systemtap/python/HelperSDT/_HelperSDT.c

I can't seem to find where -flto=auto could come from within the systemtap
sources, is that in your env somehow? That shouldn't be used when compiling
with clang. If you need a workaround, setting
CCC_OVERRIDE_OPTIONS="x-flto=auto" should work.

Later versions of clang will also ignore -flto=auto:
https://github.com/llvm/llvm-project/commit/1628486548420f85b3467026d54663d1516404f5

-- 
You are receiving this mail because:
You are the assignee for the bug.