Re: [Chicken-hackers] [PATCH] Go back to sleep when thread-join! is called without timeout.

[Peter Bex]

On Tue, Dec 10, 2013 at 12:49:09PM +0100, Christian Kellermann wrote:
> Hi!
> 
> Thanks to Michael Greenly I have the opportunity to fix a bug in
> thread-join!. Please see the commit message for details.

Hi!

Thanks a lot for the patch, and the bugreport!  I've pushed it, after
a few small modifications:

- Added the new file tests/srfi-18-signal-test.scm to distribution/manifest
- Updated NEWS
- Converted []-style parens to ()
- Reworked tests output a little so it's a bit cleaner (pending #1080)
- Removed the tricky shell gymnastics in favor of a simpler setitimer()
   call.  This should make it easier to make it work on Windows when
   (if?) we figure out a way to send signals.
- Added a cond-expand check for mingw32 which stops the program on
   Windows if we're on msys (which uses runtest.sh).

Regarding that last point, I discovered too late that my other change
for the second-to-last point broke the build anyway.  So I've had to
push a second commit to fix that.

Strangely, your way of sending the signal from the shell (via a separate
shell fork which sleeps and calls kill(1)) seems to be supported by MSYS,
and a quick test on a trivial other program seems to indicate that it
_should_ work but somehow the program never sees the signal in MSYS.
It looks like it holds the execution of the backgrounded srfi-18 test
program until the sleep and wait have finished (which is exactly what we
*don't* want, of course).

Cheers,
Peter
-- 
http://www.more-magic.net