Hi Tim, On Fri, Jul 20, 2018 at 01:40:28PM +0200, Tim Düsterhus wrote: > Am 20.07.2018 um 10:42 schrieb Willy Tarreau: > > This one is very interesting because depending on the compiler version > > the unlikely macro differs and may never detect the failure by never > > returning < 0! It's very unlikely to meet this case though. > > In fact this already caused a crash: > http://git.haproxy.org/?p=haproxy.git;a=commit;h=45be38c9c7ba2b20806f2b887876db4fb5b9457c > > You might want to review all the `unlikely` / `likely` invocations, now > that it happened twice.
You're right. I've just checked with coccinelle after a few trial and errors (I'm terribly bad at it) and managed to spot only these two ones, which is not a guarantee that it's OK but at least we don't seem to have other occurrences using the same format. Willy
