This is the mail archive of the
java@gcc.gnu.org
mailing list for the Java project.
Re: gcj with non-4k page size
- From: Thiemo Seufer <ths at networkno dot de>
- To: David Daney <ddaney at avtrex dot com>
- Cc: GCJ <java at gcc dot gnu dot org>
- Date: Sat, 27 Oct 2007 18:51:15 +0100
- Subject: Re: gcj with non-4k page size
- References: <20071026134505.GA29176@networkno.de> <472214B7.5070900@avtrex.com> <20071026164855.GB29176@networkno.de>
Thiemo Seufer wrote:
> David Daney wrote:
> [snip]
> >> The package builds fine on machines with 4k pagesize. I wonder if you
> >> ever tried other page sizes, or if this is a known bug.
> >>
> >>
> > It looks like you were doing a build with make -jx where x was not 1
> > (probably 3).
>
> I was 4 (in a 4-core system).
> The succeeding build used 1 (with a UP kernel).
>
> > Compiling HTML_401F takes a very large amount of memory
> > (and time). I wonder if the out-of-memory-killer killed it. That would
> > be consistent with the displayed symptoms. jc1 gets killed by the OOM
> > killer and prints no diagnostics, then the gcj driver program reports
> > the error you see.
> >
> > How much memory is on the system? How much swap? All of the objects it
> > looks like it was simultaneously compiling (HTML_401F, javax/swing, and
> > javax/swing/plaf/basic) cause the compiler to use a lot of memory.
> > HTML_401F uses close to 500MB and can take 20-30 minutes to compile.
>
> 1G RAM, and accidentially no swap. (The swap layout is incompatible
> between kernel pagesizes, so it was disabled.)
>
> I run now a test build on the machine with swap enabled.
FYI, this worked fine. Thanks for the hint where to look. :-)
Thiemo