On Tue, 05/28 12:01, Richard W.M. Jones wrote: > On Tue, May 28, 2013 at 11:35:20AM +0100, Richard W.M. Jones wrote: > > I'm continuing to investigate. > > Some more data points: > > v6 patch, with my laptop plugged directly into the gigabit ethernet > switch which is connected to the web server: > > - Worked perfectly 5 times in a row. > > v6 patch, with my laptop next to the wifi aerial: > > - Bug is harder to reproduce, maybe only happens 50% of runs. > > v6 patch, with my laptop about 100' from the wifi aerial: > > - Bug reproduces on every run. > > So something to do with long latency links. > > Question: Is there a place in the patch we could put a sleep in order > to simulate a long latency link? > > I also checked the logs on the web server. There are no errors, and > each access returned a 200 or 206. So the problem doesn't appear to > be at the web server end. > > Rich. >
There seems no easy way to me to inject sleep to io reqs in curl.c. Another option might be configure & compile qemu with CFLAGS=-DDEBUG_CURL and grab the stdout, so that we can see how requests are processed. -- Fam
