On Fri, Nov 09, 2007 at 07:46:53PM -0800, Christopher Layne wrote: > On Fri, Nov 09, 2007 at 06:23:33PM -0500, Nick Mathewson wrote: > > On Fri, Nov 09, 2007 at 06:27:04AM -0800, Christopher Layne wrote: > > > [ Warning: this is long and detailed, but includes details of a present > > > bug in libevent. ] > > > > > > Hi, Christopher! This all seems very reasonable to me. Would it be > > hard to turn change your proof of concept code into a regresssion > > test? If not, I'd really appreciate that, as it would make applying > > your patch a no-brainer. > > > > Yrs, > > -- > > Nick Mathewson > > I went ahead and added regression tests for this. I also cleaned up the > patch to fix a couple of things pointed out by Scott Lamb and simplified > the conditional compilation business to use the same named struct member. > Also built with "#define HAVE_SIGACTION 0" and all is good there.
Thanks, and thanks for going through the extra effort! It's much appreciated. I've checked this patch into trunk. > BTW: Should we change the regress test to not exit(1) immediately on > failure and instead exit with error at the end if any test failed? I think that would be better, yes. The only risk would be that a failure in an earlier test might make a later test give false results, but I think we could avoid that by doing the conversion carefully. Also, there are some C testing frameworks out there that people like. I don't have any experience with CUnit myself, so I can't speak to its suitability, but it's something to consider. Then again, the perfect is the enemy of the good; any patch to make this better would, well, make it better. :) yrs, -- Nick Mathewson
pgpsVGVCU4Kdv.pgp
Description: PGP signature
_______________________________________________ Libevent-users mailing list Libevent-users@monkey.org http://monkeymail.org/mailman/listinfo/libevent-users
