From: DJ Delorie <dj AT delorie DOT com>
Subject: Re: expect in b20a4?
23 Oct 1998 19:08:28 -0400
Message-ID: <36310C6C.41C6@delorie.com>
References: <Pine DOT SUN DOT 3 DOT 93 DOT 981023171753 DOT 22932D-100000 DOT cygnus DOT cygwin32 DOT developers AT modi DOT xraylith DOT wisc DOT edu>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
X-Mailer: Mozilla 3.0 (X11; I; IRIX 5.3 IP22)

Mumit Khan wrote:
> Good news is that expect is almost good enough to run the gcc/g++/etc
> dejagnu testsuites; bad news is that I'm getting segmentation violation
> after running 1700 tests or so on the average. Anybody seen it? I won't
> be able to debug it until next week, so hopefully it'll be fixed by
> then ;-)

I've successfully run "make check" on our entire devo tree without
fault or hang.  I don't know which of my changes are in the latest
beta; Geoff would know.  There were changes to both winsup and
expect, and I might have touched tcl's tclWinPipe.c as well,
and a new dejagnu/baseboards/cygwin.exp that disables some gdb tests.

The last change I made was in expect/exp_event.c in pipe_thread(),
which was to make select() time out in order to check EOF on the
pipe.  If you have this version (that pipe_thread *also*
calls PeekNamedPipe()), you probably have all my patches,
if B20 includes expect at all (Geoff?)

Before my patches, I'd also get faults after 1500-1600 gcc tests.
It's due to a number of subtle resource leaks that eventually
result in a Create*() call failing, but it isn't checked.