cURL / Mailing Lists / curl-library / Single Mail

curl-library

Re: probs. building libcurl w/gssapi on win32

From: David Rosenstrauch <darose_at_darose.net>
Date: Thu, 22 May 2008 15:00:06 -0400

Dan Fandrich wrote:
> On Thu, May 22, 2008 at 08:00:51PM +0200, Yang Tse wrote:
>> You must have skipped the part where you have to #define
>> _KRB5_WIN_MAC_H 1 before including _any_ GSSMIT header in libcurl, as
>> this is what should prevent the inclusion of the contents of GSSMIT's
>> <win-mac.h> even when it is included no matter from where.
>>
>> The effect must be to first get undefined symbols, not duplicate ones,
>> when building libcurl.
>>
>> Once you get that going is the moment to include only the symbols
>> actually needed in "libcurl_gssmit_win_mac.h".
>
> This approach might work, but it ties libcurl dangerously close to the
> internals to GSSMIT. I'd much rather see an approach that sidesteps
> this altogether.

Please note - no offense intended here. I really am trying to help.
But please be aware: I spent time creating and submitting a patch for
this. I then spent more time completely redoing the patch the way Yang
requested. Now you're saying that's wrong.

I'm happy to try to help here, but I don't have oodles of time to spend
on this. Can you guys please come to an agreement soon as to how you
want this to be patched and then let me know? Or if you can't, then can
one of you please take over creating this patch?

Just my $0.02 on a solution:

The number of curl users who build from scratch is not that big, as a
percentage of the total number of curl users. Ditto for the number of
curl users on Windows. And ditto for the number of curl users who use
it with MIT GSS. But the intersection of those 3 sets is extremely
small - possibly even just 1 single person: me! IMO, that's how this
code has managed to be broken for so long (I see these compile errors in
v7.15.4 - and I'm even wondering if this code has always had compile
errors on Windows). So given that there's such a small number of people
impacted by this, it seems to me that it's not something that should
warrant major amounts of debate - or major changes to the source code -
to fix. So IMO, just try to find this simplest thing that fixes this
and be done with it. Again, just my $0.02, though.

DR
Received on 2008-05-22