Re: [ANNOUNCEMENT] TEST RELEASE: Cygwin 2.3.0-0.6

[Corinna Vinschen <corinna-cygwin@cygwin.com>]

[-- Attachment #1: Type: text/plain, Size: 2149 bytes --]

On Nov  5 10:24, Corinna Vinschen wrote:
> On Nov  4 22:25, Achim Gratz wrote:
> > Achim Gratz writes:
> > > Just confirmed this on another system.  If I kill the child, then the
> > > parent resumes and finishes the test loop alright and it can be
> > > interrupted again from the shell.  The hang happens after the first test
> > > succeeds in both the parent and child.
> > 
> > I have just managed to kill the parent (returning the shell prompt) and
> > have the child complete the test loop output to the terminal.  So I
> > guess the communication ping-pong is somehow buggered up so that pipes
> > start blocking.
> 
> Staring into the latest version of my new function to remove pending
> signals, after having some *more* coffee, it seems pretty clear I
> screwed this up nicely.
> 
> What I was missing all the time was to iterate over the list of pending
> signals if there's a pending signal which doesn't have to be cleared.
> This case was just missing.  Duh!  I guess I didn't really cover myself
> in glory here...
> 
> I applied yet another patch and uploaded a new developer snapshot
> (this time *with* the ACL changes) to https://cygwin.com/snapshots/
> 
> Can you please give it a try ASAP?

For the records, I got a testcase from Achim to reproduce the issue
which, incidentally, is still present in that snapshot.  It turned out
that the problem Achim was reporting has nothing to do with the new code
clearing pending signals.  It was a completely different bug, which just
showed another behaviour due to the signal change.

In fact what happened was a deadlock when parent and child process
were industriously trying to lock a file using flock(2).  This must
have been present for a long time, in fact.

I (hopefully) fixed the issue and uploaded yet another developer
snapshot to https://cygwin.com/snapshots/ (with ACL changes again).

New test release follows soon.

Achim, thanks for the report and the testcase.


Corinna

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

[-- Attachment #2: Type: application/pgp-signature, Size: 819 bytes --]