Stefan wrote on Sun, Feb 07, 2016 at 20:51:08 +0100:
> On 2/7/2016 01:22, Daniel Shahaf wrote:
> >Stefan wrote on Mon, Feb 01, 2016 at 00:34:32 +0100:
> >>+++ fs-loader.c (working copy)
> >>@@ -461,7 +461,8 @@
> >> if (! svn_utf__cstring_is_valid(path))
> >> {
> >> return svn_error_createf(SVN_ERR_FS_PATH_SYNTAX, NULL,
> >>- _("Path '%s' is not in UTF-8"), path);
> >>+ _("Path '%s' is not in UTF-8"),
> >>+ path ? path : "NULL");
> >Is this actually a problem? svn_error_createf() uses apr_pvsprintf()
> >which (by code inspection) accepts NULL here.
> TBH I didn't look further... Just assumed that apr_pvsprintf() would just be
> a redefine. If it's a specific implementation which checks for NULL here and
> handles it, then pls disregard my patch.
https://svn.apache.org/viewvc/apr/apr/tags/1.3.0/memory/unix/apr_pools.c?revision=661875&view=markup#l1100
https://svn.apache.org/viewvc/apr/apr/tags/1.3.0/strings/apr_snprintf.c?revision=661875&view=markup#l953
I didn't check later versions. (In theory, apr-2.0 might have
changed this behaviour. I have no reason to think it did.)
> P.S. I tried to check out the function definition myself, but could only
> spot the declaration in the apr source. Couldn't trace down where the
> function is defined...
I found it with ctags(1) with -IAPR_DECLARE:
% ctags --extra=+qf --c-types=+p --python-types=-i -IAPR_DECLARE -R .
% grep apr_pvsprintf tags
apr_pvsprintf include/apr_strings.h /^APR_DECLARE(char *)
apr_pvsprintf(apr_pool_t *p, const char *fmt, va_list ap);$/;" p
apr_pvsprintf memory/unix/apr_pools.c /^APR_DECLARE(char *)
apr_pvsprintf(apr_pool_t *pool, const char *fmt, va_list ap)$/;" f
% vim -t apr_pvsprintf
Cheers,
Daniel
