libxul compilation problem

[Robert Bonomi]

> From owner-freebsd-questions at freebsd.org  Sun Oct 17 11:46:48 2010
> Date: Sun, 17 Oct 2010 18:47:09 +0200
> From: =?ISO-8859-1?Q?Fernando_Apestegu=EDa?= <fernando.apesteguia at gmail.com>
> To: Rob Farmer <rfarmer at predatorlabs.net>
> Cc: User Questions <freebsd-questions at freebsd.org>
> Subject: Re: libxul compilation problem
>
> 2010/10/16 Rob Farmer <rfarmer at predatorlabs.net>:
> > 2010/10/16 Fernando Apestegu=EDa <fernando.apesteguia at gmail.com>:
> >> I didn't run X or whatsoever. That's why I think I should have enough me=
> mory.
> >> In fact after getting that error, I rebooted so I could update the
> >> ports from a "fresh"
> >> running system (nothing cached or so). But even in that case, I'm gettin=
> g the
> >> same error.
> >>
> >> Any VM tuning I can try?
> >
> > I'm not really knowledgeable about that kind of thing.
> >
> > However, the port is marked MAKE_JOBS_SAFE which means that it will
> > try to run multiple compiler instances in parallel, to speed things up
> > if you have multiple CPUs/cores. You can try running with "make
> > -DDISABLE_MAKE_JOBS" to just run one at a time - maybe you have enough
> > memory for that but not multiple jobs at once?
>
> Hi Rob,
>
> The machine has one single core cpu. Finally I was able to compile the
> thing, compiling
> the offending file by hand (nsHtml5ElementName.cpp) without the -O2
> optimization flag.
> With this flag, cc1plus eats up all the memory of my system in a few
> seconds. Without
> the flag, the file is compiled without any problems and quite fast.
>
> Should this issue be a candidate for filing a PR?

*ONLY* if you can provide a 'fix' _with_ the report!  <grin>
(Make sure the fix works on a machine with only 64mb ram and 256m swap. )

Turning on optimization virtually _always_ results in the compiler needing
more resources.   "How much" more depends on the size, complexity, and '
optimizability' of the code being compiled.

The "simple" fix for your problem is to add swap space to the system.
swap space does -not- have to be in a dedicated partition, see 'man swapon'
for how to use a -file- as temporary swap space.


Note: if you find someting that won't compile, given a combined 4 gigs of
RAM and swap space, and the build isthe only thing running beyond core
system services, *then* you've got the basis for 'good' PR filing.