Bug#846520: pcl: enable parallel builds on 32 bit arches
On 01/12/16 23:25, Jochen Sprickerhof wrote: > Hi Emilio, > > * Emilio Pozuelo Monfort[2016-12-01 22:24]: >> Why was this disabled? Do the builds run out of memory often? It'd be good >> to re-enable this, the speedup is huge and it would avoid taking one buildd >> for a day and a half on some architectures. > > Exactly because it was running out of memory. Quoting from the buildd: > > | cc1plus: out of memory allocating 903468 bytes after a total of 19152896 > bytes > > I asked in #debian-buildd back then and the consensus was to remove the > --parallel. Is there any reason to revert this now? We have more powerful buildds, at least in some architectures. Not sure about RAM/swap. Also there was one bad build back then, but several good builds. I'm not sure if this is such an issue, but I think it's worth re-trying. Doesn't need to happen right now just for the sake of it, it can be uploaded when there are other worthwhile changes. If parallel support is re-enabled and some builds fail, we can evaluate whether it's best to disable parallel support on specific architectures, blacklist pcl in specific buildds, lower the parallelism without disabling it entirely, or disabling it altogether in 32-bits architectures again. Cc'ing debian-wb-team@ to see if anybody disagrees. Thanks, Emilio -- debian-science-maintainers mailing list debian-science-maintainers@lists.alioth.debian.org http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/debian-science-maintainers
Bug#846520: pcl: enable parallel builds on 32 bit arches
Hi Emilio, * Emilio Pozuelo Monfort[2016-12-01 22:24]: > Why was this disabled? Do the builds run out of memory often? It'd be good > to re-enable this, the speedup is huge and it would avoid taking one buildd > for a day and a half on some architectures. Exactly because it was running out of memory. Quoting from the buildd: | cc1plus: out of memory allocating 903468 bytes after a total of 19152896 bytes I asked in #debian-buildd back then and the consensus was to remove the --parallel. Is there any reason to revert this now? Cheers Jochen signature.asc Description: PGP signature -- debian-science-maintainers mailing list debian-science-maintainers@lists.alioth.debian.org http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/debian-science-maintainers
Bug#846520: pcl: enable parallel builds on 32 bit arches
Source: pcl Version: 1.8.0+dfsg1-3 Severity: wishlist Hi, In your changelog, I see: pcl (1.7.2-6) unstable; urgency=medium [ Jochen Sprickerhof ] * Remove --parallel from dh to fix build on i386 buildd. -- Leopold Palomo-AvellanedaTue, 02 Dec 2014 08:29:43 +0100 My guess is that happened because of the build failure on 2014-11-30 in: https://buildd.debian.org/status/logs.php?pkg=pcl=i386 Note how the builds on i386 take ~5 hours, while on amd64 they take ~1h, because of the parallel builds. For e.g. mipsel it's at 22h or 1d15h vs. 8h on mips64el. Why was this disabled? Do the builds run out of memory often? It'd be good to re-enable this, the speedup is huge and it would avoid taking one buildd for a day and a half on some architectures. Thanks, Emilio -- System Information: Debian Release: stretch/sid APT prefers unstable APT policy: (800, 'unstable'), (700, 'experimental'), (650, 'testing'), (500, 'unstable-debug') Architecture: amd64 (x86_64) Foreign Architectures: i386, armhf Kernel: Linux 4.7.0-1-amd64 (SMP w/4 CPU cores) Locale: LANG=en_GB.utf8, LC_CTYPE=en_GB.utf8 (charmap=UTF-8) Shell: /bin/sh linked to /bin/dash Init: systemd (via /run/systemd/system) -- debian-science-maintainers mailing list debian-science-maintainers@lists.alioth.debian.org http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/debian-science-maintainers