[Slackbuilds-users] openjdk8 compile (either manually or via sbotools) failing at openjdk7

Arnaud arnaud.garcia-fernandez at laposte.net
Thu Apr 7 20:19:03 UTC 2022


On Thu, 7 Apr 2022 22:17:56 +1000
Christoph Willing <chris.willing at linux.com> wrote:
> 
> I don't normally have /tmp as tmpfs and the build has consistently
> worked OK.
> 
> I just tried with /tmp as tmpfs and had a few failures because of
> insufficient "disk" space. The reason is quite explicitly explained when
> it happens so I don't think that is OP's problem.
> 
> Eventually the build succeeded when I had 10G allocated (it used ~8.3G)
> and the build was quite a bit faster (23mins vs 36mins).
> 
> chris

Well, you're right, I just rebuilt it on a tmpfs place.
It is faster, and doesn't wear SDD drives, and my box has 32G of RAM,
hence I'm pretty OK with a 24G tmpfs for building stuff.

So eliminating process shows :
 - tmpfs or SSD are both OK.
 - JOBS=1, JOBS=5 or JOBS=9 doesn't change the outcome.
 - setting LANG=C, LANG=en_US.UTF-8 or LANG=fr_FR.UTF-8 is the same (we
never know when unicode might break something old !)
 - sourcing /etc/profile.d/gcc5.sh matters not.
 - multilib chroot or normal chroot are both working. Today. Because
yesterday's multilib chroot wasn't working (it had fontforge,
webcore-fonts, and some python packages for Kallithea and buildbot, I
removepkg'ed all non-slackware and non-multilib stuff).


So I think I figured it out finally, and we should have these sentences
added to the README :
«
  This build may consistently fail.
  In case it happens, please do unrelated stuff, sleep it off, and retry
  another day.
  It might pretty well consistently work by that time.
 »

	- Yth...


More information about the SlackBuilds-users mailing list