Stas Bekman wrote:
Randy Kobes wrote:
On Mon, 12 Jul 2004, Stas Bekman wrote:
Randy Kobes wrote:
I still have it crashing ...
Please try the following 3 variations (the first one should certainly work as it doesn't call cleanup).
Unfortunately, only the 1st one works for me ... The 2nd one crashes after completing two tests, and the 3rd one after one test.
That's what I thought. But I really need to be able to reproduce it to be able to find out what the problem is :(
Philippe, where are you with this $bb->cleanup issue?
Well, things are indeed strange. I've just updated httpd an hour ago and I can still reproduce the problem here, on my work machine. At home, on my other almost identical linux box (both Fedora Core 2), I can't get it to happen. Very strange indeed. The current stacktrace I get is exactly the same, and the corrupted memory values are still the same.
(gdb) print *b
$1 = {p = 0x8512858, list = {next = 0x1, prev = 0x1011007}, bucket_alloc = 0x861cd28}
(gdb) print *(b->bucket_alloc)
$2 = {pool = 0x0, allocator = 0x0, freelist = 0x0, blocks = 0x0}
(gdb) print *(b->p)
$3 = {parent = 0x8681394, child = 0x8, sibling = 0x13, ref = 0xb, cleanups = 0x58,
allocator = 0xff1,
subprocesses = 0xc3a5e50, abort_fn = 0xc3a3dd8, user_data = 0x0, tag = 0x0, nodes =
0xa84bdb4,
file_line = 0xc3a4964 <Address 0xc3a4964 out of bounds>, creation_flags = 32,
stat_alloc = 0,
stat_total_alloc = 176471476, stat_clear = 205146472, owner = 64, mutex = 0x0}The bucket allocator exists, but seems to be NULLed, and that can't be a good sign. The pool in the bb also looks somewhat corrupted.
Apart from this, I don't quite have the time to trace thru this at this time.
Thoughts ?
BTW, didn't we also have a problem with ARP::Pool::cleanup on win32? I remember disabling those tests, as I couldn't reproduce it on my machine. It seems to be related.
Yes, we did - there's a block of tests in TestAPR/pool.pm
that are commented out. I just tried these, and now they
all pass for me! So I guess we can re-enable them?
Can you check your original report and see whether you now have a different setup? I don't think I've changed anything around APR::Pool, so if it suddenly started to work, it's not a good sign.
Actually, it might be interesting to see if Philippe (who also had difficulty with the echo_bbs2 tests) has any problem with these re-enabled pool tests.
Nope, on this system :
Failed Test Stat Wstat Total Fail Failed List of Failed ------------------------------------------------------------------------------- t/protocol/echo_bbs2.t 3 2 66.67% 2-3 2 tests skipped. Failed 1/188 test scripts, 99.47% okay. 2/1261 subtests failed, 99.84% okay.
Yeah, Philippe, can you help us here? Or help me reproduce the segfaults?
-- -------------------------------------------------------------------------------- Philippe M. Chiasson m/gozer\@(apache|cpan|ectoplasm)\.org/ GPG KeyID : 88C3A5A5 http://gozer.ectoplasm.org/ F9BF E0C2 480E 7680 1AE5 3631 CB32 A107 88C3A5A5
signature.asc
Description: OpenPGP digital signature
