[169] in Pthreads mailing list archive
Re: oops, spoke too soon.
daemon@ATHENA.MIT.EDU (Christopher Provenzano)
Thu Oct 26 15:25:03 1995
To: "William S. Gribble" <grib@cs.utexas.edu>
Cc: pthreads@MIT.EDU
In-Reply-To: Your message of "Wed, 25 Oct 1995 15:13:49 CDT."
<199510252013.PAA22557@cascade.cs.utexas.edu>
Date: Thu, 26 Oct 1995 14:53:17 EDT
From: Christopher Provenzano <proven@MIT.EDU>
> I posted earlier today that beta-4 was crapping out on my test code.
> Just about every problem I had has been resolved:
>
> 1. I said sys/socket.h ``worked with beta3'' and I wondered why the change.
> answer: it didn't work with beta-3 either, and I had forgotten that I fixed
> it locally.
Unfortuanately, I don't have access to enough Linux systems to ensure that
pthreads works on all of them. I compiled it right out of the box on a 1.2.x
system and a 1.3.x system but because there are so many versions of the C
libarary out there this doesn't always work.
>
> 2. I said signals weren't delivered between threads or from outside the
> process. True, but the fix I posted to pthreads-bugs a while back
> still applies. Line numbers have changed substantially, but the
> patch still goes into sig_handler_fake() in signal.c.
>
> 3. I said my regression tests all failed. That was because I had added
> some code to my PThread class to try to get larger stack sizes, and had
> forgotten to remove that code when I determined that stack size wasn't
> the problem. The fact that changing the stack size attribute doesn't work
> is another problem; but removing the stack-size-modifying code from the library
> I posted makes the regression tests work fine.
>
> Sorry to be so alarmist.
>
> Thanks,
> Bill Gribble
I'll be looking into the other problems soon.
