Re: Dead locks in t/runtests
On 2011-08-18 21:24, Niels Thykier wrote:
> Hi
>
> I have been working a bit on debugging the dead-lock issue in t/runtests
> and I have found something interesting. Unfortunately I cannot
> pin-point the issue...
>
> When I see a dead-lock it, I noticed that we tend to have two lintian
> proceses (ps: [lintian] <defunct>) and two /usr/bin/sort. Do we know
> that IPC::Run/Lintian::Command is thread-safe?
>
> ~Niels
>
>
As you may have noticed, I have decided to follow that hunch and removed
all (read: "both") uses of Lintian::Command/IPC::Run in t/runtests.
So far I have not experienced any dead-locks. It is not a proof that
it now works, but I am very pleased with apparent sudden absence of
dead-locks ,,, even when running with 16 threads (on a related note; the
runtime does not really improve after 8 threads). :)
~Niels
Reply to: