pthreads update for the adventurous

["Robert Collins" <robert dot collins at itdomain dot com dot au>]

This is a followup to the long thread early this week about g++
exception handling crashing with a sigsegv.

I've implemented the necessary functions to build the cygwin gcc
2.95.3-2 source tarball with posix thread support. The resulting gcc,
when building the test case (using sched_yield as that forced the fault)
does not crash or fault in any way (that I could see).

Snapshots with a date greater that 2001-Apr-10 have the new
functionality. The new functionality includes all of the pthread_*
functions from the single unix specification with two exceptions:
pthread_atfork & static initializers. Static initializers for mutex's
exist, and will be introduced for other objects in due course.

Cancellation point functions are present, but at the moment are just
stubs: cancellation points will require a substantial amount of work
across all of the cygwin library functions... so they'll be coming in
later.

This is a plea for testing. On this list we've seen discussions about
pthreads for
g++,
gcj,
miniORB,
python
and who knows what else in the past.

If you are regularly rolling your own with any of those three
applications, please test the latest snapshot (available anytime now)
and report back any errors or hiccups. I suspect the appropriate list
for feedback is the cygwin-apps list - I'm posting here because most of
the discussions took place here.

Special note: g++ testing requires rebuilding gcc to take advantage of
threads. ONLY DO THIS IF YOU"VE DONE IT BEFORE. I'm not familiar with
the innards of gcc and thus not able to help you out of any problems
that may occur.

There is some minor urgency on this: I would really like to hear back on
any issues with the pthreads code before Chris freezes cygwin and rolls
1.3.0 - which is only a few days away.

Finally, IIRC GCJ was foundering on some sort of threads-related issue.
It'd be great if someone could look into that with the new snapshot.

Rob


--
Want to unsubscribe from this list?
Check out: http://cygwin.com/ml/#unsubscribe-simple