[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [csmith-dev] Any tricks for narrowing down GCC bugs to avoid duplications?



Hi Arthur,

This isn't off topic at all!

But we don't have a good answer, unfortunately. Not sure about my colleagues, but I've spent quite a bit of time thinking about this problem without much success.

One idea is to do a binary search over GCC versions and look for the first rev that contains a given bug. If two failures with identical symptoms were introduced by the same commit, it seems likely they are duplicates. But I tried this and it wasn't that helpful: too many times, the first-broken-versions for two failure-inducing test cases were different but the bugs were the same anyway.

Your bug report is good.  But it may be a duplicate of this one:

  http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48694

Given the large number of open bugs it can be very hard to spot these.

Your bug report is better than mine because it has a reduced test case! I used to try to reduce resource hog bugs but it's so painful that I stopped. So nice job :).

The one thing I'd do differently is include the "gcc -v" output, which may contain some useful detail.

We have a lot of work-in-progress on automated testcase reduction. We hope to release it in the not-too-distant future. But if you are interested in being an early tester, let us know and we can set you up.

John