This is the mail archive of the cygwin-developers@sourceware.cygnus.com mailing list for the Cygwin project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: w32api divergence issues


Chris Faylor <cgf@cygnus.com> writes:
> On Mon, Nov 01, 1999 at 01:46:01PM -0600, Mumit Khan wrote:
> >The w32api headers are diverging away from Anders' set, and that's going
> >to start causing problems unless we do regular merges back and forth.
> >
> >My source tree is based on Anders' 0.2.0 snapshot from 1999-08-02, and
> >then some minor local changes. This is quite different from what's in
> >winsup right now, and I really hope that these changes have been
> >forwarded to Anders.
> 
> I haven't looked at Anders header files recently.  The last time I
> corresponded with him, he was talking about changing the include files
> to use configure.

The version in Winsup is 0.1.5 snapshot + a few more changes that Anders
had sent in later. There were the other changes directly in winsup of
course (eg., NET* changes by Corinna, an OLE header fix by DJ, few by
you and others), but nothing major.

In the meantime, Anders had published a 0.2.0 snapshot, incorporating
lots of fixes and changes that had been reported, and I switched over
to that.

My current tree is based 0.2.0, merged in the changes from winsup (and
fixed 2 problems in winsup headers, both added later and I don't know
why!), and also quite a few changes to COM/OLE and a few other places.

Feel free to take a look at my 1999-11-07 snapshot (source + binary):

  ftp://ftp.xraylith.wisc.edu/pub/khan/gnu-win32/cygwin/gcc-2.95.2/

> Also, IIRC, the newer header files actually broke cygwin in some strange
> way.

I remember that. Some of those were simply bugs in w32api headers 
(missing pragma pack is one prominent cause), and some may have been 
bugs elsewhere (gcc, winsup, etc).

> I'll certainly investigate a patch if someone wants to send one.

Just heard from Anders after a long time saying that his priorities are 
a bit different now ("just moved again"), and will update on his 
situation; that means that we shouldn't wait for him to make the move. 
I could create a diff between mine and winsup and submit that of 
course, and perhaps a few volunteers could build winsup and run for a 
few days to check??

I'm open to suggestions; I'd hate to see a big divergence, which is what
had happened in windows32api days until Geoffrey did a massive merge one
time (see ChangeLog for Oct 19, 1998). It gets very messy merging headers.

> We are talking about setting up a repository but I'm advocating that
> we should really think about consolidating gcc/gdb/cygwin/binutils into
> one coherent repository.  Otherwise we'll have one more CVS tree which
> includes, for instance, libiberty.

The workaround for now may be to also tar up CVS repo for winsup/newlib 
just like the snapshots. I've done this in the past to avoid security 
issues of setting up remove repos, and did work for what I had intended 
it for.

Regards,
Mumit


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]