Re: firefox on i386: the return of EFBIG

[Paul Ripke <stix%stix.id.au@localhost>]

On Sat, Apr 06, 2024 at 08:06:31AM -0500, John D. Baker wrote:
> On Fri, 5 Apr 2024, David Brownlee wrote:
> 
> > On Wed, 3 Apr 2024 at 16:08, John D. Baker <jdbaker%consolidated.net@localhost> wrote:
> > >
> > > Although it seemed to have been fixed as of netbsd-10_RC2, and recent
> > > bulk-build reports don't reflect this problem, I am once again having
> > > trouble building "www/firefox" from pkgsrc-2024Q1 on NetBSD/i386-10.0
> > > (no bloody RC1, 2, 3, 4, 5, or 6).
> > 
> > Are there any other differences between the working netbsd-10_RC2 and
> > problem NetBSD/i386-10.0 system? I assume nothing obvious like PAE or
> > running compat32 on an amd64 kernel :)
> 
> No, it's the same setup.  amd64 hardware running a real i386 netbsd-10
> with a PAE-enabled kernel.  The only glitch I've seen was that sometime
> in early March I accidentally updated with i386-current over the -10
> install.  This caused 'postinstall fix obsolete' to delete some -10
> libraries, but I re-extracted the -10 sets and manually removed the
> conflicting libraries (minor version higher than in -10).
> 
> The build host normally runs 9.3_STABLE/amd64 from local disk.  When
> building for -9/i386, -10 or -current (i386 or amd64) it runs as a
> semi-diskless client (local disk only for swap and scratch space).

I hit the same error building firefox on netbsd-10 aarch64. It seems an
8MiB stack ulimit is insufficient; bumping to 16MiB allowed the build
to complete. Note that I haven't actually tested the build - it's on
a free Oracle VM :)

-- 
Paul Ripke
"Great minds discuss ideas, average minds discuss events, small minds
 discuss people."
-- Disputed: Often attributed to Eleanor Roosevelt. 1948.