Re: INTERNAL: Exiting with 2 jobserver tokens available; should be 5!

[Jaak Ristioja]

Hi!

On 13.11.2016 07:37, Tim Murphy wrote:
> Something like Valgrind might spot some initial problem that doesn't
> immediately crash but eventually spirals out of control.

I could try valgrind, but (1) I will need to recompile glibc with debug
symbols to use it, and (2) I don't have an eternity to wait for
--trace-children to finish, so I'd have to run it without
--trace-children because afaik valgrind doesn't provide good means to
trace only certain children (i.e. only "make" processes). Actually the
RPi2 would probably OOM/die from multiple valgrind processes anyway.

> I don't know what the gcc version is on your Pi but if you have a recent
> enough one  you might manage to use the address sanitiser option to get
> a similar result.

I'm currently using GCC 5.4, so its fairly new from that aspect, but I
won't be able to use its address sanitizer, because it doesn't work with
a PaX/grsecurity kernel like Gentoo's sys-devel/hardened-sources, due to
"ASAN assumes/uses hardcoded userland address space size values, which
breaks when UDEREF is set as it pitches a bit from the size" [1].
Because of this, it is disabled by default on hardened profiles, hence
I'd have to both recompile GCC and a kernel without PaX/grsec to

So I guess I'll attempt to recompile glibc and run valgrind on the
parent "make -j5" process and see whether that turns up anything. If
not, then I'll try the -fsanitize=address approach. I expect all of this
to take some time (and perhaps wear out more flash storage) on the slow
RPi2.

Best regards,
J


[1] http://blog.siphos.be/2013/04/another-gentoo-hardened-month-has-passed/