"Wilson, Douglas" <dgwilson@sonomasystems.net> writes:
> The following are the patches I used to get Subversion to compile on AIX
> 4.3.2
> with 'CC=xlc' and 'CFLAGS='-g -qfullpath -qdbxextra'
> (--enable-maintainer-mode didn't
> seem to be good enough for debugging). Sadly, most of the tests still fail.
> I believe
> this is because of a broken (or incompatible?) iconv. I tried updating my
> iconv package
> (those that show up with 'lslpp -l | grep iconv', BTW, the README.aix file
> in the latest perl
> distribution is a great help for those who need to do this and are not
> expert AIX
> sysadmins); anyway, that didn't help either. Does anyone know of any other
> related packages
> to update?
>
> I tried configuring with '--without-iconv' but that gets ignored in
> apr-util. I see there's
> an apr-iconv project in process which I may try if I can figure out how to
> plug it in
> to apr-util.
Do you think you have broken translators, or no translators at all.
If it is the former (broken translators) then if you hacked the
functions get_ntou_xlate_handle and get_uton_xlate_handle in
libsvn_subr/utf.c so that they both always set *ret to NULL then no
conversion to or from UTF8 would be attempted. You might have
compatibilty problems with Subversion on systems that do have UTF8,
but the majority of the regression tests should then work. At least
it would confirm that the UTF8 stuff is the problem.
>
> patches work with '-p 6' from the subversion-r3200 directory.
>
> *** /cvs/src/tst/orig/subversion-r3200/subversion/libsvn_subr/io.c
> 2002-09-20 13:52:55.000000000 -0700
> --- /cvs/src/tst/patched/subversion-r3200/subversion/libsvn_subr/io.c
> 2002-10-17 09:41:07.000000000 -0700
> ***************
> *** 528,534 ****
> (apr_err, 0, NULL, pool,
> "svn_io_file_checksum: error reading from '%s'", file);
>
> ! apr_md5_update (&context, buf, len);
>
> } while (! APR_STATUS_IS_EOF(apr_err));
>
> --- 528,534 ----
> (apr_err, 0, NULL, pool,
> "svn_io_file_checksum: error reading from '%s'", file);
>
> ! apr_md5_update (&context, (const unsigned char *) buf, len);
Ah yes, the char/unsigned char problem. I think gcc warns about these
conversions if we use -Wcast-qual.
--
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Oct 19 20:38:43 2002