Bug 629342

Summary:	LTO regression in gcc-6x (and above) causes build failures on gparted and inkscape
Product:	Gentoo Linux	Reporter:	Steve Arnold <nerdboy>
Component:	Current packages	Assignee:	Gentoo Toolchain Maintainers <toolchain>
Status:	RESOLVED OBSOLETE
Severity:	major	CC:	tsmksubc
Priority:	Normal
Version:	unspecified
Hardware:	All
OS:	Linux
URL:	https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69953
See Also:	https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69953
Whiteboard:
Package list:		Runtime testing required:	---
Bug Depends on:
Bug Blocks:	618550

Description Steve Arnold archtester

2017-08-30 01:51:31 UTC

Unfortunately the upstream bug is closed (again) and marked RESOLVED/FIXED but it seems the patch (backported to at least gcc 6.4.0) has reintroduced it (too much reduction maybe?).

Both apps were building/working with gcc 5.3/4 but the same error now occurs with said patch.  The root cause is essentially one group of upstream devs not using (or misusing) proper visibility rules for public vs private symbols, and another group of upstream devs deciding how the compiler should handle it for different levels of optimization (obviously they haven't converged).  Add to that the un-specificity of the language standard and this is what you get.

I can't comment on the upstream bug or I would reopen it (plus I have travel this week) so I guess I'm documenting it here.  Maybe there's a vapier patch in there somewhere...

I can add my current gparted build failure here, but it's exactly the same as before (and documented on that gcc bug).

Comment 1 Andreas K. Hüttel archtester

2017-09-01 21:15:10 UTC

@Steve: please add all the info here as needed on the upstream bug (preprocessed sources) and I'll forward it.

Comment 2 Steve Arnold archtester

2017-09-06 19:46:58 UTC

Pastbin okay for now?

https://paste.debian.net/984852/

plus I have no idea what code to add, since it's a linker failure.  Can you be more specific?

Comment 3 Andreas K. Hüttel archtester

2017-09-06 20:00:46 UTC

(In reply to Steve Arnold from comment #2)
> Pastbin okay for now?
> 
> https://paste.debian.net/984852/
> 
> plus I have no idea what code to add, since it's a linker failure.  Can you
> be more specific?

I attached your log upstream and asked for advice there. Let's see.

Comment 4 Steve Arnold archtester

2017-09-28 23:22:42 UTC

One more data point; i got gparted to actually build with the "same" config and versions of the toolchain on a different machine; the hardware is newer and the Gentoo piece is a little bit more up-to-date (and possibly more libs rebuiilt with  newer toolchain bits).  I tried a quick rebuild of the main deps and then gparted but still the same fail (on the original machine).  More analysis required...

Comment 5 Sergei Trofimovich (RETIRED) gentoo-dev

2019-02-14 07:35:25 UTC

Does it happen on stable gcc-7.3.0? If it does can you post emerge --info and build.log?

Comment 6 Sergei Trofimovich (RETIRED) gentoo-dev

2019-07-17 22:24:36 UTC

Closing as obsolete: not reproducible locally. No `emerge --info` posted to attempt to reproduce closer to the problematic environment.