This isn't the right fix:
Index: lib/query.c
===================================================================
RCS file: /v/rpm/cvs/rpm/lib/query.c,v
retrieving revision 2.215.2.2
retrieving revision 2.215.2.3
diff -p -u -w -r2.215.2.2 -r2.215.2.3
--- lib/query.c 22 Dec 2010 22:39:04 -0000 2.215.2.2
+++ lib/query.c 8 Jan 2011 20:31:46 -0000 2.215.2.3
@@ -692,12 +692,13 @@ int rpmQueryVerify(QVA_t qva, rpmts ts,
}
qva->qva_mi = rpmtsInitIterator(ts, RPMTAG_FILEPATHS, fn, 0);
- if (qva->qva_mi == NULL && !provides_checked && !gotpattern)
+ if (qva->qva_mi == NULL && !provides_checked && !gotpattern) {
qva->qva_mi = rpmtsInitIterator(ts, RPMTAG_PROVIDENAME, fn, 0);
#if defined(RPM_VENDOR_MANDRIVA)
if(rpmmiCount(qva->qva_mi) == 0)
qva->qva_mi = rpmmiFree(qva->qva_mi);
#endif
+ }
if (qva->qva_mi == NULL) {
struct stat sb;
Well yes, you need the curlies in the right place, but
you should NOT have to call rpmmmiCount() to detect
whether a Provides: is not present.
(aside)
The real flaw is that RPM needs to FORBID insanities like
Provides: /path/to/file
Looking in multiple indices, and in multiple databases, just because
some idiot decided to smoke some crack about
Provides: /path/to/file
and invent stories why Newer! Better! Bestest! is needed. It is NEVER
correct to add data values in multiple places.
73 de Jeff
______________________________________________________________________
RPM Package Manager http://rpm5.org
Developer Communication List rpm-devel@rpm5.org
