Anton Pevtsov wrote:
> Martin, unfortunately, these changes results in segmentation fault both
> on Windows and Unix.
> I'll investigate this, but could you take a look on it too when you have
> a chance?
Sure. I tested (as usual) with gcc and EDG eccp on SPARC/Solaris
and saw no problems, but I have just reproduced the crash on IA32
Linux. It turns out to be an incompatibility between the signature
of the test function and the type it's being cast to in the driver
(in _rw_dispatch). Fixed thusly:
http://svn.apache.org/viewvc?rev=410398&view=rev
Martin
|