On Thu, Mar 13, 2008 at 12:33:30PM -0500, Craig A. Berry wrote: > With HP C V7.1-015 on OpenVMS Alpha V8.3, -Duseithreads, I get > > Failed 4 tests out of 1080, 99.63% okay. > [-.lib.ExtUtils.t]cd.t > [-.lib.ExtUtils.t]eu_command.t > [-.lib.ExtUtils.t]xs.t > op/threads.t > > Fixes are available for cd.t and xs.t; since these are test patches > only and will not affect what gets installed, is it kosher to just > patch them and not take on the risk of a full-scale integration?
Well, I was hoping to get the Win32 ExtUtils::Install writable directories issue sorted, and then merge everything in lib/ExtUtils across in one hit, as it all seems to be stable. > eu_command.t does not fail when run by itself and I haven't figured > out what it's tripping over (possibly a temp file left by some other > test that gets cleaned up later). Hmm. > op/threads.t also fails on 5.8.8 (not the version that shipped with > it, but the new version of the test fails when run on 5.8.8). The test > that fails is labeled: > > [perl #45053] Memory corruption with heavy module loading in threads > > and it may well be that our dynamic loading code is a bit lacking in > the thread safety department. The crash is below. If this is the > only thing holding up 5.8.9, I think we should document it as a known > problem and move on. I don't recognise it, so documenting it seems to be the best idea. Nicholas Clark
