On 2017/10/19 13:49, Sebastien Marie wrote:
> On Thu, Oct 19, 2017 at 12:23:29PM +0100, Stuart Henderson wrote:
> >
> > Hi - this failed on i386,
> >
> > [...]
> >
> > Full log from that build attached. I'm trying it again with
> > RUST_BACKTRACE=1 in MAKE_ENV to see if that gives more, but since
> > it's rust I don't expect any results for some hours ;-)
>
> RUST_BACKTRACE=1 is useless on OpenBSD as the platform doesn't support
> full backtrace.
>
> could you check the size of the coredump ?
>
> it could be a memory problem (limit reached): it failed to build
> librustc.so which is the part which require the most memory to achieve.
>
> my i386 buildhost has 3Go of RAM available for building (but I didn't
> check the exact limit required for rust).
>
> thanks.
> --
> Sebastien Marie
It built OK in rebuild, and dpb cleaned the build directory, so I can't
check the core size any more.
Building with 2 dpb cores on a 2-cpu-core machine so there wouldn't have
been many other processes running - but if a build is very tight on memory,
it fails randomly (presumably due to PIE?) - I've seen that often with
chromium before I started using -Wl,--no-keep-memory.
$ ulimit -d
3145728
