Justin Erenkrantz jerenkrantz@apache.org writes:
This patch removes a bunch of compiler warnings seen with Sun Forte.
Besides one 'statement not reachable' warning, the rest are
signed/unsigned conflicts. The real root of this stems from the fact
that apr_md5_* only deals with unsigned char*'s while everything else
in SVN only deals with char*s. (Does GCC ignore these warnings?)
Perhaps the bigger question is whether to somehow fix the apr_md5_*
API to take char*'s instead? I'm sort of on the fence here. Before
starting what will most likely be a painful discussion on dev@apr,
I'd like to see what you guys (namely Karl since he wrote most of the
checksum code that uses MD5) think first. What makes more sense?
I have to admit, it makes more sense to me to fix APR. I mean, most
callers are probably passing signed 'char *' data. If APR's md5 code
internally needs to treat it as unsigned, for the purpose of the
digest update algorith, it should just cast.
No, I never saw any warnings about it, not from GCC 2.96 anyway.
Justin, would you be willing to have that discussion on dev@apr? I'll
bet most callers are either casting, or not doing anything because
they're compilers aren't warning them anyway... But I haven't looked
at (say) httpd.
(For those who didn't see the patch, note that we're not talking about
the retured digest here, but about the input data -- normally some C
string buffer. The digest is always unsigned, of course.)
-Karl
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Oct 14 02:04:32 2006