Ben Collins-Sussman wrote on Thu, 15 May 2008 at 11:46 -0500:
> OK, I've reviewed the branch -- not the KDE-specific code, of course,
> but the general auth-system changes it entails. Arfrever has
> responded to my review with fixes, but there's only one issue I'm
> still worried about.
>
> Specifically, I don't understand why this change is creating an
> entirely new svn_auth_kwallet.h public header file. As a rule, we try
> to keep the number of public headers files to a minimum, to keep
> things simpler for developers. This new header seems like a wart to
> me; all it does is declare a new kwallet auth provider. And yet, our
> other "conditional" auth providers (crytpostore on Windows, keychain
> on OS X) are declared directly in svn_auth.h already.
>
> Arfrever said that when his provider was declared directly within
> svn_auth.h, that danielsh said it "broke the windows build" in IRC.
> Can we have some more investigation on that problem, please, rather
> than just turn a blind eye and create a new header? :-)
>
To clarify: whenever svn_auth_get_kwallet_simple_provider was declared
in svn_auth.h -- even if its declaration was inside a "#if 0" block --
Visual Studio added it to the symbols list in 'libsvn_subr.def' and
tried to compile a 'kwallet.obj' file. (The libsvn_subr.def file is
generated by VC, not by the Python scripts.)
The build error message and the output from "gen-make.py --debug" are at
http://paste.lisp.org/display/60584. Using Visual C++ 9 Express.
Daniel
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe_at_subversion.tigris.org
For additional commands, e-mail: dev-help_at_subversion.tigris.org
Received on 2008-05-15 19:59:53 CEST