This is the mail archive of the cygwin mailing list for the Cygwin project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: proc_retry question


On Dec 11 09:02, xmoon 2000 wrote:
> On 8 December 2014 at 16:27, Corinna Vinschen <corinna-cygwin@cygwin.com> wrote:
> > On Dec  8 16:03, xmoon 2000 wrote:
> >> On 8 December 2014 at 15:50, Corinna Vinschen <> wrote:
> >> > On Dec  8 14:48, xmoon 2000 wrote:
> >> >> On 8 December 2014 at 14:40, Corinna Vinschen <> wrote:
> >> >> > On Dec  8 12:40, xmoon 2000 wrote:
> >> >> >> I have set proc_retry = 10.
> >> >> >>
> >> >> >> I still get the following error message but showing rety 10:
> >> >> >>
> >> >> >>
> >> >> >>       0 [main] sh 35392 fork: child -1 - forked process 35736 died
> >> >> >> unexpectedly, retry 10, exit code 0xC0000005, errno 11
> >> >> >> /q/onlyPastEnd.sh: fork: retry: Resource temporarily unavailable
> >> >> >>
> >> >> >> Does this mean it got an error and cygwin will now retry 10 times?
> >> >> >>
> >> >> >> OR does it mean it retry-ed 10 and failed?
> >> >> >
> >> >> > It tried 10 times and failed.
> >> >>
> >> >> Are you certain about that?
> >> >
> >> > No.  Sorry, it's the other way around.  If retry is 10, no retry
> >> > occured.  AFAICS that's because you get an C000005 status code, which is
> >> > STATUS_ACCESS_VIOLATION.  In case of this error code, Cygwin does
> >> > explicitely not retry.
> >>
> >> Do this suggest that I need to do the rebaseall - as documented in:
> >> http://x.cygwin.com/docs/faq/cygwin-x-faq.html#q-fork-failures
> >
> > You can try, but this looks a bit different.  A SEGV at fork time could
> > be a bug in Cygwin or a result of having another Cygwin DLL in the path
> > for some reason.
> 
> I tried "rebasing" but with no luck.
> 
> The problem appears when I am running 60 scripts in parallel on a 32
> core machine. Plus, each script runs subscripts with several pipes.
> So, I think this is a "too many processes" type of issue.

It's a SEGV.  It's probably not related to having too many processes.
Cygwin processes only care for their immediate child processes usually
and there's a restriction to 256 or so.

> Usually, windows handle my "over-coring" well and my machine just
> works through what it needs to do. But when I try this from Cygwin, it
> seems to cause issues.
> 
> Any thoughts or suggest for how I could get around this?

Apart from BLODA influence, or apart from debugging what causes the
SEGV, no.  If you can provide a simple testcase, stripped to the bare
minimum of code to reproduce the issue, it may help.


Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Maintainer                 cygwin AT cygwin DOT com
Red Hat

Attachment: pgp8lR537mSlp.pgp
Description: PGP signature


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]