[Patch] Update for the testsuite, devdsp

Corinna Vinschen vinschen@redhat.com
Tue Aug 24 09:58:00 GMT 2004

On Aug 21 15:39, Pierre A. Humblet wrote:
> At 02:13 PM 8/17/2004 +0200, Corinna Vinschen wrote:
> >On Aug 16 23:09, Pierre A. Humblet wrote:
> >> This is the first time I run the testsuite, and it was on WinME. 
> >> There were more failures than I expected, e.g. in mmap. I don't
> >> know how this compares to NT.
> >
> >I guess it might make sense to discuss this on cygwin-developers.
> >Could you send your results?  Chris and I both ran the testsuite on
> >XP and the FAIL count is 0.
> Attached are winsup.sum and winsup.log
> There are many mmap failures (including a dump in
> mmaptest01), I have not looked at them.

I did now.  I shouldn't have.  It makes bad dreams. 

mmaptest01 crashes because of this extraordinary dumb problem on 9x.
When creating two maps using MapViewOfFile which are mapping the same
address in the same file, the function always returns a pointer to the
same shared region.  The testcase relies on mmap returning two different
addresses in adjacent mmap calls.  It unmaps one of them, which results
on 9x in munmapping both.  The following access to this memory page
then crashes on 9x.  Since I don't see a way to workaround that in
Cygwin, I've changed the testcase to accomodate that behaviour.

The FAILs in mmaptest02 and mmaptest03 are expected FAILs since the
tests rely on the VirtualProtect ability, which 9x doesn't have on
shared pages.

> ltp/symlink01 has a SIGSEV, but I think it happens during a printf
>  triggered by a hard link related failure.
>  So it's either the test program or newlib. 

The test program uses %d format specifiers for printing inode numbers.
I changed the testcase to use %llu instead.


Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Project Co-Leader          mailto:cygwin@cygwin.com
Red Hat, Inc.

More information about the Cygwin-developers mailing list