On Thu, May 7, 2009 at 10:14 PM, Colin Watson <[email protected]> wrote: > The original problem that led me to find this bug was reported here: > > https://bugs.launchpad.net/ubuntu/+source/findutils/+bug/373214 > > ... and amounts to "/usr/include/asm/ contents missing from Linux kernel > package when built with findutils 4.4.1". > > > The -prune action needs stat information when -depth is not given, but > find's parser fails to declare this. It is therefore possible for -prune > to use undefined stat data. > > I've attached a patch (against the rel-4-4-fixes branch, I'm afraid; I > have spent the whole afternoon trying to track this down in the face of > half our developers complaining because all their builds are breaking, > so I'd rather do something else right now, but I wanted to get this > patch sent; if you need it reconstructed against master then I can do > that tomorrow),
That's OK, I can do it. > including a test I can't tell you how grateful I am that you actually included a test :) Does the test reliably detect a problem if the bugfix is not present? Is the "oldfind" binary affected too (all tests are routinely run against both binaries at all valid values of find's -O option) ? > that you may be able to get to > demonstrate the problem. Whether the test actually fails for you is > unfortunately not certain, since it's an undefined-data problem; you > should at least be able to see the bug manually by breaking on > pred_prune. It's possible there's something we can do to make a failure more likely or more reproducible if similar problems occur in the future, for example setting variables to 'invalid' values before calling fts_read. Whatever we do we will need to make sure it does not break fts. Thanks for the patch! James.
