Bug 29362

Summary:	gcc dies while trying to build qtwebengine5 on Mageia 8
Product:	Mageia	Reporter:	David Walser <luigiwalser>
Component:	RPM Packages	Assignee:	Thomas Backlund <tmb>
Status:	RESOLVED INVALID	QA Contact:
Severity:	normal
Priority:	Normal
Version:	8
Target Milestone:	---
Hardware:	All
OS:	Linux
Whiteboard:
Source RPM:	gcc-10.3.0-1.mga8.src.rpm	CVE:
Status comment:
Bug Depends on:
Bug Blocks:	29249

Description David Walser 2021-08-12 05:24:54 CEST

The same code has built multiple times without issue in Cauldron, but it has died both times I've tried in Mageia 8, so I doubt it's the build system going homicidal on it.

g++: fatal error: Killed signal terminated program cc1plus
compilation terminated.

Comment 1 David Walser 2021-08-12 05:25:38 CEST

Full build log (both attempts):
http://pkgsubmit.mageia.org/uploads/failure/8/core/updates_testing/20210812015132.luigiwalser.duvel.13908/log/qtwebengine5-5.15.5-1.mga8/build.x86_64.0.20210812015233.log
http://pkgsubmit.mageia.org/uploads/failure/8/core/updates_testing/20210812015132.luigiwalser.duvel.13908/log/qtwebengine5-5.15.5-1.mga8/build.x86_64.0.20210812015233.log

David Walser 2021-08-12 05:59:44 CEST

Blocks: (none) => 29249

Comment 2 Thomas Backlund 2021-08-12 14:53:04 CEST

the chromium code build in qtwebengine is a memory hog that manages to kill gcc due to running out of memory...

building it locally here without cpu restrictions used up 64GB of ram, and filled up 32GB of swap before dying...

I limited it to max 4 cpus for now to get it through the buildsystem...

Seems Cauldron toolchain is more resilient to this memory usage than the mga8 toolchain... so a subtle bug somewhere in mga8 toolchain

Comment 3 David Walser 2021-08-12 15:08:57 CEST

No wonder it almost killed my machine when I tried to build it locally.  I only have 16G of RAM.  I just ordered more.

Comment 4 Thomas Backlund 2021-08-12 17:52:43 CEST

 
build succeded

Comment 5 David Walser 2021-08-12 22:24:00 CEST

Thanks!!!

To be fair, digikam failed in Cauldron for the same reason, so I guess we're hitting our heads against the ceiling with heavy C++ stuff again; no big surprise.

Resolution: (none) => INVALID
Status: NEW => RESOLVED