httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "William A. Rowe, Jr." <wr...@covalent.net>
Subject Re: cvs commit: httpd-2.0/server core.c
Date Thu, 03 Jan 2002 02:53:47 GMT
From: "Bill Stoddard" <bill@wstoddard.com>
Sent: Wednesday, January 02, 2002 8:25 PM


> > BTW --- 1.3.23 question here; The API_EXPORT -> API_EXPORT_NONSTD changes in
> > theory require an MMN bump on Win32.  Which really sucks, because I don't want
> > to break everyone else.  Anyone given any thought as to how we can avoid an MMN
> > bump, when all that's happened is a specific platform has been broken [linkage,
> > etc?]  It almost makes me think we _aught_ to drop the .def file for win32,
> > creating decorated symbol names that now-broken modules will refuse to link
> > against, sparing all other platforms a version bump.
> 
> I -think- we can ditch the .def file. I'm a bit concerned that we might break something
> unexpected...  Also need to bump the MMN_MAJOR. Doing anything else is a kludge that
could
> introduce more problems than it solves.

I believe that this will destroy binary compatibility.  Ergo - no bump.  There is no
API change or structure change in 1.3.23 that harms compilation, or precompiled .so's
on other platforms.  Win32 was the only 'victim' of the cdecl/nonstd changes.

Since changing the ApacheCore.dll wipes out binary compatibility, in and of itself,
there should be no reason to bump the MMN.  1.3.22 and prior modules simply won't
load, and simply won't suffer from segfault-ish behavior.  It's all or nothing if
we destroy the .def, but we don't create headaches for the other platforms.  That's
the reason I suggested it.

Bill


Mime
View raw message