Re: quandary with pthreads

[Thomas Pfaff <tpfaff at gmx dot net>]

Rob,

thank you that you are willing to add my patches.

FYI, the logic of the mutex stuff is running in pthreads-win32 about 1 1/2
years now unchanged (guess who has contributed).

I will try to add some of the testcases from pthreads-win32.

I still see some problems with WIN95 since it lacks support for
InterlockedCompareExchange which is required in pthread_mutex_trylock
(and the fact that InterlockedIncrement and Decrement return only 0,
positive or negative instead of the real counter value).

I have created assembler versions for the Interlocked functions that work
fine unless they are made inline (see attachment). Maybe someone with a
better understanding of gcc inline assembler might have a look at it).

Nevertheless adding these functions will drop support for the good old
i386.

Greetings,
Thomas


On Wed, 11 Dec 2002, Robert Collins wrote:

> I *really* want to get Thomas's contribution properly reviewed in time
> for 1.3.18.
>
> Unfortunately my cygwin development machine's hard drive is failing, and
> it will be off the air for ~ 1 week.
>
> So, I'd like to suggest the following:
>
> Thomas..
> If you can write test cases for all the new capabilities you introduce -
> see the ones I added based on your previous work, and the README in the
> test-suite dir.
> Then I will be happy to have your changes committed if the test suite
> shows no pthreads regressions on a win9x machine, and on an NT kernel
> based machine.
>
> Best I can offer right now, sorry.
> Rob
>
> --
> ---
> GPG key available at: http://users.bigpond.net.au/robertc/keys.txt.
> ---
>

Attachment: winbase.h
Description: Text document

Attachment: winbase.c
Description: Text document