From: Shawn O. Pearce Date: Thu, 29 Oct 2009 14:56:04 GMT Subject: Re: [RFC PATCH v4 26/26] test smart http fetch and push Message-ID: <20091029145604.GV10505@spearce.org> In-Reply-To: <20091029143702.GU10505@spearce.org> "Shawn O. Pearce" wrote: > Junio C Hamano wrote: > > > + s/^Content-Length: .*$/Content-Length: xxxx/ > > > > This chomped line is so unlike you---what happened? > > I was getting different Content-Lengths on different runs of the > test. I don't know why. Here the Content-Length is of the gzip'd > request, it shouldn't be varying with each run, but it seemed to be. FWIW, I found the variablity here. It was triggered by running the test with -v or without -v, which caused clone to change its first "want" line to include or exclude progress messages, based on the tty-ness of stderr. Adding a --quiet to the clone command means I can avoid this munging and hard code the length in the test vector. But I'm still against doing that here in case the protocol capabilities ever change in the future. -- Shawn.