Odd failures during parallel PPC builds

[Jukka Ukkonen]

Hi all,

I sent the question below to Peter Grehan a few days ago.
He told me that he has not seen the symptoms I had found
possibly because he does ppc cross-builds on amd64.
I have tried to maintain my amd64, ppc, and sparc64 systems
as independent as possible. Usually I cross-build only my
R-PI on amd64.
Now it seems I have found something going wrong with
parallel make on ppc such that the same symptoms cannot
be reproduced on an amd64 platform.

> I just wonder am I the only one to see odd failures during
> parallel builds on a 32-bit ppc system with "make -j #" in
> which # >= 2. Has anyone else noticed similar symptoms?
> The make version is apparently bmake. The OS version I have
> used is 10-stable.
> When I use the traditional 1 process at a time approach
> with PPC builds everything comes out just fine, though,
> about 1,5 times slower on a 2 core system.
>
> On a 12 core amd64 "make -j 60" is just fine, but on
> a 2 core ppc "make -j 2" often seems to be too much.
> I have no idea of the root cause yet. I am just confused.
> This has anyhow been going on for a few weeks now. So, I
> assume someone else might have already noticed something
> similar.
> Typically the error codes during parallel builds have been
> ENOENT and EINVAL during the last few weeks. The failure
> does not happen always at the same time, nor is the error
> code always the same.
>
> Are there maybe known conditions like configuration
> settings or something which could cause these odd failed
> parallel builds on a ppc system?

Is there anyone else trying to run local builds on
32-bit ppc? If so, have you noticed any problems with
parallel make? Is there any other way to work around
this problem but avoiding the parallel builds completely?

BR,
--jau