[Zope3-dev] win32 Zope 3 Tests

[Jeremy Hylton]

>>>>> "GC" == Godefroid Chapelle <gotcha@swing.be> writes:

  GC> This implies that the fact that I am using Borland Compiler to
  GC> build extensions seems to not be a problem.

This test failed because the Borland memcmp() return value is not just
-1, 0, or 1.  This is a difference between Borland and MS on Windows,
but it's still a bug in our code.  Fixed in CVS.

  GC> Further, this implies that the multiple "connection refused" I
  GC> already mentioned in
  GC> http://lists.zope.org/pipermail/zope3-dev/2002-November/003896.html
  GC> seem also not to be related to the Borland Compiler built.

The tests work fine on Win2k and Win98 with the MS compiler, so I
don't know what else it could be.

  GC> IOW, I have the feeling that it would really be worth to examine
  GC> the problem.

  GC> I am ready to help in any ways compatible with my null
  GC> experience regarding sockets.

We really need someone with a little sockets experience to help.  My
first guess is that the problem relates to some difference between the
MS and Borland socket code, which points to some nonstandard behavior
that our tests are relying on.

When you run the ZEO tests, is it always the same tests that fail?  Do
they fail if you run them in isolation?  If you run them in a loop
(-L), do they fail every time?

Jeremy