Pre-3.6 code freeze

[Bruce Gray]

On Jul 18, 2011, at 5:40 PM, James E Keenan wrote (in reply to Coke's  
clean smoke test from Win7):
--snip--
> It seems we have a much greater variance in the files which are  
> failing in different smoke reports on Windows than we get in any  
> other OS.
> Do you have any ideas as to what enabled us to get a PASS on this  
> box as distinct from other Windows boxes?

I mis-read the question, but I think my mistaken reply is still worth  
posting:

Win32 smoke reports vary so widely because of...
different OSes, filesystems, compilers, and compiling environments:
     MSVC vs CygWin vs MinGW
     Strawberry_Perl vs ActivePerl
     Win2k vs WinXP vs Vista vs Win7
     FAT vs NTFS

For example, CygWin and MinGW use different "host" environments, and  
they both use a different `make` and link to a different C runtime  
than MSVC.
Unicode (and probably also IPv6) support varies both by filesystem and  
by OS version.

I think it is manageable long-term; it certainly *needs* to be  
managed, due to the high number of Win32 users in the world.
We just took our eyes off the ball for too long.

Several of the Win32 bugs I am pursuing simply boil down to "take this  
test that is already SKIPped in some scenario, and also skip it on  
Win32 when in FOO environment".

-- 
Hope this helps,
Bruce Gray (Util)