FYI: the response from Bill re: the .def files and APR_DECLARE.
Thanks, Bill!
Cheers,
-g
----- Forwarded message from "William A. Rowe, Jr." <wrowe@rowe-clan.net> -----
From: "William A. Rowe, Jr." <wrowe@rowe-clan.net>
To: "'Greg Stein'" <gstein@lyra.org>
Subject: RE: APR patch: Win32 fixes
Date: Mon, 6 Nov 2000 21:39:18 -0600
> From: Greg Stein [mailto:gstein@lyra.org]
> Sent: Monday, November 06, 2000 7:31 PM
>
> Is this necessary? I know that Bill Rowe was considering
> torching the .def file altogether. (I forget the reason
> why, but he mentioned this at ApacheCon)
No... we aren't torching it right now. But I will be regenerating
it, and eliminating the ordinals. The ordinals are unneeded since
MS does a quite decent job with hashed lookups of symbol names.
The .def is needed since aprlibdll creates the dll from the lib
file, not the obj's. This eliminates one more nasty issue against
binary compatibility.
> Is this needed because we aren't using APR_DECLARE()
> throughout the APR header files? e.g. to get it exported,
> this change or the header change needs to happen?
The cool bit of APR_DECLARES ... when we get to another alpha
cycle, I'm building a dll straight from the .objs - all the
APR_DECLAREs will advertise what is exported. Based on that
export list, we have the 'true' symbol list. This can be, if
I've calculated right, piped right back into a common symbol
list for all the platforms. The other side of the APR_DECLAREs,
(even with the current structure that looses the symbols when
building dll from lib) is that the fixups are correct and more
optimized than a general case, and their is a single fixup
rather than a list of fixups throughout the .obj.
----- End forwarded message -----
--
Greg Stein, http://www.lyra.org/
Received on Sat Oct 21 14:36:14 2006