Stephan Beal wrote:

> On Tue, Aug 19, 2014 at 7:31 PM, Warren Young <> wrote:
> > On 8/18/2014 19:39, Eric Rubin-Smith wrote:
> >
> >>
> >> warning: Can't read pathname for load map: Input/output error.
> >>
> > That looks like a corrupted stack to me.
> > Try running it under Valgrind.
> +1

I ran it under valgrind already and mentioned the results in the OP -- 
you probably missed that in my wall of text :-).  The upshot is that 
valgrind's imposed slow-down causes us not to reach the bad path.  So no 
help there.

If you follow my post through to the end, though, you will see that 
the call to 'xfer_accept_file' in xfer.c line 1641 is strongly
implicated by my gdb session.  If you go and look at that function,
you will see a call to blob_extract on line 139 that passes 'n', where
'n' was obtained above from the incoming data (xfer.c line 130).  It seems 
very likely that we are placing too much trust in this 'n' during 
'blob_extract'... though this is just my suspicion of course, not a 

> and while i'm here: thank you for the tremendously detailed bug report.

no problem.  I've been on the receiving end of tickets like "teh
software doesn't work 4 me can u plz fix it" enough times to try to 
avoid inflicting similar pain. :-)

Eric A. Rubin-Smith

fossil-users mailing list

Reply via email to