Dan Price wrote: > On Mon 09 Jun 2008 at 05:54PM, [EMAIL PROTECTED] wrote: > >>> I wish there was an easy way of reproducing the timeout conditions locally. >>> >> The way I've been testing this is to insert calls to sleep() at >> [in]opportune moments in the depot code. It's helped me catch a lot of >> stupid edge cases. >> >> -j >> > > Perhaps some bit of code is needed so we could plumb this sort of > delay-injection testing into the test harness? > > I'm not quite sure what that might look like, though. One option > would be to have a new "stress" section of the test suite which one > could run... continuously. >
I don't know if it would help in this case, but somewhere in Sun there is a kernel module designed specifically to create artificial latency, dropped packets, etc. for testing for software errors or performance issues in these kinds of situations. It's called "hitbox". I'd have to do some searching, but if someone knows the author perhaps they could be convinced to release it for this kind of testing? - Matt _______________________________________________ pkg-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
