On 6/25/20 7:08 PM, Ján Tomko wrote:
On a Wednesday in 2020, Laine Stump wrote:
The only reason for the error label in this function is to call virBufferFreeAndReset(). It's actually more common for a failed format function to just leave the virBuffer alone and let the caller free it when there is a failure, and in fact the only caller of this function that *wasn't* already calling virBufferFreeAndReset() on failure was virDomainDefFormat() (via virDomainDefFormatInternal()).
qemuDomainDefFormatXMLInternal does not call it either.
Dang! I thought I had followed every call chain with cscope, but maybe I just searched in this one file? Anyway, it's especially embarrassing because not only did I miss qemuDomainFormatXMLInternal(), I also missed virDomainSnapshotDefFormat (which called virDomainSnapshotDefFormatInternal(), which calls virDomainDefFormatInternal()) :-( I think as a followup patch, I should convert every occurrence of "virBuffer blah = VIR_BUFFER_INITIALIZER" to "g_auto(virBuffer) blah = VIR_BUFFER_INITIALIZER" - in a quick search just now I already found a couple more (totally unrelated to virDomainDefFormat) that aren't properly cleared out on error. Thanks for taking the time to actually fact check my claims. #FakeCommitLogs
That is easily solved by modifying virDomainDefFormat() to declare its virBuffer buf with g_auto(), so that virBufferFreeAndReset() is automatically called.
Signed-off-by: Laine Stump <laine@redhat.com> --- src/conf/domain_conf.c | 88 ++++++++++++++++++++---------------------- 1 file changed, 42 insertions(+), 46 deletions(-)
With that fixed: Reviewed-by: Ján Tomko <jtomko@redhat.com>
Jano