Re: sitecopy hangs downloading forever

On Tue, Oct 04, 2005 at 10:46:33AM -0500, Timothee Besset wrote:
> Joe Orton wrote:
> Actually, I need to update what I said above: the local file keeps 
> growing forever, or at least as long as I left it grow. The remote file 
> is 1.1Gb, and after leaving it running overnight, my local file was at 
> 1.8Gb and still growing.

I just did a --synch of a ~1.7 Gb file using DAV over SSL here without 
problems (over the loopback device, anyway).  What server are you using, 
what version?  Can you determine whether the server reports the correct 
content-length for the large file; e.g.:

  curl -k --head https://server.com/path/to/the/big/file

> We are removing the bigger files and gonna transfer them some other way. 
> I can't really run the tests above as I need the bandwidth to perform an 
> actual transfer that will work for sure today. The strace log would be 
> really huge, and I don't think the http one would show anything other 
> than the endless 'got 8196 bytes in response body'.

Well, if the process is stuck doing system calls that would be useful to 
know, so that I can see *what* system calls.  If it's not stuck doing 
system calls that also would be useful.

> This looks very much like an int overflow. I attached gdb to the process 
> once it had gone past the actual size, and couldn't get a sensible 
> backtrace. Possibly at SSL level if the whole file is transfered with a 
> single call into SSL? ( I had 3-4 frames of SSL stuff, everything above 
> was garbled ).

gdb backtraces can be grabled depending on stack setup at interrupt 
time; e.g. try "break ne_read_response_block" then "cont".

Or even enable debugging then;

(gdb) set ne_debug_mask=4095
(gdb) cont

and the debug output will be enabled.

There's not much I can do about the problem without some further 
diagnosis, in any case, sorry.

Regards,

joe
_______________________________________________
sitecopy maillist  -  [email protected]
http://dav.lyra.org/mailman/listinfo/sitecopy