Andres Kroonmaa wrote:
> oh, I meant more all the ping-ponging between clientSendMoreData
> and storeClientCopy with weird seemingly random buffer allocs
> and frees as needed on the way.
The same buffer is reused until it is freed..
> I only mentioned this piece because I've had several backtraces
> passing through these lines. Sure I might be mislead. I gotta stare
> more at that module source before I can really talk ;)
If we are seing the same bug, then it is quite unlikely that the
stacktrace indicates where the problem is. The malloc problem has most
likely been introduced some time before, in perhaps a completely
different routine. The fact that the stacktraces are different, all
crashing inside malloc is further evidence of this.
The use of a good malloc debugger without mempools should spot this
immediately, provided a situation where the problem can be reproduced is
found without requiring a huge workingset of allocations... (the problem
with good malloc debuggers is that they are slow, and quite often bloat
the memory usage by many orders of magnitude)
/Henrik
Received on Fri Nov 10 2000 - 13:00:08 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:12:57 MST