glibc detected *** python: free(): invalid pointer: 0x00007fe3d443e829

[obluda]

hey dude,
i've encountered some freeing probs.
i've checked add_complete method and i think you're trying to free too often in the for loop. also not freeing anything incase that for was not being ran at all.

*** glibc detected *** python: free(): invalid pointer: 0x00007fe3d443e829 ***
======= Backtrace: =========
/lib/libc.so.6(+0x71e16)[0x7fe3d57c2e16]
/lib/libc.so.6(cfree+0x6c)[0x7fe3d57c7b8c]
/lib/python2.7/site-packages/pyreBloom.so(add_complete+0x54)[0x7fe3d443e534]
/lib/python2.7/site-packages/pyreBloom.so(+0x4cb6)[0x7fe3d443fcb6]
python(PyObject_Call+0x36)[0x44a176]
/lib/python2.7/site-packages/pyreBloom.so(+0x40ca)[0x7fe3d443f0ca]
python(PyEval_EvalFrameEx+0x9c6)[0x4eb166]
python(PyEval_EvalFrameEx+0xa82)[0x4eb222]
python(PyEval_EvalCodeEx+0x1a0)[0x4f1db0]
python[0x4f2f32]
python(PyObject_Call+0x36)[0x44a176]
python(PyEval_EvalFrameEx+0x1fb9)[0x4ec759]
python(PyEval_EvalCodeEx+0x1a0)[0x4f1db0]
python[0x4f2f32]
python(PyObject_Call+0x36)[0x44a176]
python[0x47cc7a]
python(PyObject_Call+0x36)[0x44a176]
python(PyEval_CallObjectWithKeywords+0x36)[0x44c666]

[obluda]

oops, i gave it a second look and all seems fine..
mm weird.

[dlecocq]

I've seen this happen very occasionally, but I'm not quite sure why. It hasn't gotten in the way of it being useful for what we use it for yet, but it's on our radar.

[obluda]

i've been hunting it with debug version and valgrind, weird things were happening on hiredis level.
then i realized i have some really archaic hiredis library..
everything works ok so far ;-)

also, what do you think about freeing bloom.c:96:reply = redisCommand(ctxt->ctxt, "PING"); ?

[dlecocq]

Blerg -- I should have seen that. Good catch! I'll patch that.

[obluda]

i've also noticed segfaults when redis server restarts. to reproduce:
1/connect with pyre
2/restart redis
3/do some add request

[widnyana]

hi @dlecocq,

pardon me, did you have patch for this?

[dlecocq]

I just tried locally the steps provided by @obluda , and that doesn't reproduce the behavior for me. @widnyana -- do you have any steps to reliably reproduce this behavior?