Bug 241030 - build timeouts for powerpc64
Summary: build timeouts for powerpc64
Status: Closed FIXED
Alias: None
Product: Ports & Packages
Classification: Unclassified
Component: Package Infrastructure (show other bugs)
Version: Latest
Hardware: powerpc Any
: --- Affects Only Me
Assignee: Port Management Team
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2019-10-03 09:05 UTC by Piotr Kubaj
Modified: 2019-10-12 16:36 UTC (History)
2 users (show)

See Also:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Piotr Kubaj freebsd_committer 2019-10-03 09:05:13 UTC
There are numerous build timeouts on powerpc64 that happen in the cluster (for gcc-arm-embedded, libreoffice and various llvm versions).

I already reported them via email to portmgr, but either they didn't receive those emails or for some other reason didn't change timeouts.

Examples:
http://pylon.nyi.freebsd.org/data/head-powerpc64-default/p513097_s352826/logs/errors/llvm-devel-10.0.d20190913.log
http://pylon.nyi.freebsd.org/data/head-powerpc64-default/p513097_s352826/logs/errors/llvm80-8.0.1_3.log
http://pylon.nyi.freebsd.org/data/head-powerpc64-default/p513097_s352826/logs/errors/gcc-arm-embedded-8.2.20181220_1.log
Comment 1 Baptiste Daroussin freebsd_committer 2019-10-03 11:18:57 UTC
I am really puzzled about that. 30h is already very very long, the solution is probably to request fresh new hardware for the cluster.
Comment 2 Piotr Kubaj freebsd_committer 2019-10-03 11:45:27 UTC
(In reply to Baptiste Daroussin from comment #1)
The hardware is AFAIK Tyan machines with POWER8, so not that slow. I think it's more about ppc64 itself being slower than amd64 on FreeBSD (the situation is improving thanks to ongoing work by some developers).

AFAIK the Foundation also bought some Talos systems, but I don't know when they will be ready for work in the cluster (sbruno probably knows more).

Anyway, those packages build in Poudriere on powerpc64 on my own server, but I can't give you the time that is necessary, because I have ALLOW_MAKE_JOBS=yes set.
Comment 3 Piotr Kubaj freebsd_committer 2019-10-12 16:36:30 UTC
Seems fine for now, looks like setting 40h was a right choice (rust and paraview took more than 30h, they would have failed with old settings).