On Thu, May 3, 2012 at 10:47 AM, Chad Rosier <[email protected]> wrote:
> > On May 3, 2012, at 10:37 AM, Chandler Carruth wrote: > > On Thu, May 3, 2012 at 10:07 AM, Chad Rosier <[email protected]> wrote: > >> Author: mcrosier >> Date: Thu May 3 12:07:55 2012 >> New Revision: 156079 >> >> URL: http://llvm.org/viewvc/llvm-project?rev=156079&view=rev >> Log: >> [driver crash diagnostics] Remove more flags that reference absolute >> paths that >> aren't necessary to reproduce the clang crash. >> > > Out of curiosity, is it easy to test this by triggering a crash in the > test suite, and then FileCheck-ing the resulting repro stuff? > > > In r155205 (http://llvm.org/viewvc/llvm-project?view=rev&revision=155205) > I added a simple hack to cause the compiler to crash on demand. We could > add a test case to cause a crash, but what were checking is the contents of > the generated script file, which has a non-deterministic name. You'd have > to grep the name from the clang output and then FileCheck the output > script. Doesn't sound too difficult, but since this is off the critical > path I honestly haven't spent a lot of time trying to get that to work. I > can do so if you (or whoever) really think it's necessary. > It's not a big deal... I was more thinking if we're going to keep tweaking this, maybe we should get some testing coverage of it. If you're interested in testing it, I would just add an environment variable much like the one you added to force the crash (maybe the same one?) that forces the output to a particular location. Actually, now that I think about it, I have a very distinct use case for exactly such an environment variable... maybe I can write up a patch and send it your way?
_______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
