Eric Blake wrote:
According to Jim Meyering on 2/19/2010 3:45 AM:
Hum, that one I'm not sure. In the case of virBufferGrow failure, we just did va_end(locarg); in the loop before, so going to cleanup there does it twice, and I'm not sure it's legal. Probably simpler to add just va_end(argptr); before return in that case and drop the cleanup: target.
Good catch. Corrected, as you suggest:
+++ b/.gnulib @@ -1 +1 @@ -Subproject commit 11fbc57405a118e6ec9a3ebc19bbf5ececdae4d6 +Subproject commit a1d565aa5b47ec56243e74d4ac8b4988a751fef2
Shouldn't that be an independent patch?
Yes. BTW, if I were to update to the latest from gnulib, it would break "syntax-check", due to a weakness in maint.mk's new hash.h check. It gets a false-positive on any inclusion of libvirt's own "hash.h": $ git ls-files|grep hash.h src/util/hash.h I haven't yet decided what to do about that.