Re: Core dump (Illegal instruction) with crosstool's GCJ

[Martin Egholm Nielsen <martin at egholm-nielsen dot dk>]

Hi,

> > > Maybe Martin should consider running the full set of gcj
> > > regression tests.  It's the only way to go when trying out the
> > > compiler on a poorly-supported platform.  I certainly did this
> > > when I started using gcc on ppc405, but I never got around to
> > > doing it for gcj on ppc405.
> > I may try so, yes... Haven't looked into that at all - is it easy
> > to set up?
> It is not nearly as easy as building the toolchain, and it requires 
> significant effort afterwards to understand the results and
> communicate them to the gcc team (and rerun tests with various
> patches applied as time goes on), but if you're serious about wanting
> gcj on that platform, it's well worth it.
I think I'll get the time for it within the next couple of days/weeks...

> Also, you might have a platform setup problem.  One of the first 
> steps in running the tests the way I do it is to set up a chroot jail
> populated entirely with the shared libraries built by crosstool, plus
> the shell from the target (or one built using crosstool). See
> http://www.kegel.com/crosstool/current/doc/chroot-login-howto.html 
> Once you have the jail working, you can run the tests remotely using
> crosstest.sh as a guide.  Note that I have never run the gcj 
> regression tests.  Also note that it's much, much easier to run the
> tests natively if possible; I did it remotely because my ppc405
> didn't have enough RAM or disk to do that. - Dan
Well, I have 32 megs of RAM and the root-filesystem mounted over NFS, so 
unless that amount of memory is not sufficient, I'll think it'll work...

Best regards,
 Martin



------
Want more information?  See the CrossGCC FAQ, http://www.objsw.com/CrossGCC/
Want to unsubscribe? Send a note to crossgcc-unsubscribe@sources.redhat.com