[Xapian-devel] Re: [XapianBug 117] Invalid data in remote protocol can cause buffer overflow

[Olly Betts]

On Thu, Apr 05, 2007 at 03:01:26PM +0100, Olly Betts wrote:
> On Thu, Apr 05, 2007 at 02:32:40PM +0100, James Aylett wrote:
> > apitest on remote backend hangs as well (on emptyquery1,
> > the first one that tries to do anything), but remotetcp runs fine. I
> > guess there's a problem with the prog backend on Mac OS X latest?
> 
> There's something up with a change I committed earlier today I think.
> I'm currently investigating.

The issue I'm seeing turns out to be unrelated to my change - I stepped
back through SVN some way and revision 7901 failed in the same way (I
didn't try earlier than that).

The issue is that a build with assertions enabled (mine has all enabled,
but I've not tried just "partial" yet) when run under valgrind (so I
suspect James' issue is different) just stalls on each test until the
timeout happens.  The command I'm running is this (from the "tests"
subdirectory):

    ./runtest ./apitest -b remote

I've had a poke around, and I can't see what's up.  My guess is that
there's a consistency check in the remote backend somewhere which
runs really slowly under valgrind, or that valgrind's emulation of
some syscall the remote backend uses is flawed.  If anyone else has
a similar build on a machine with valgrind, it would be interesting
to know if it behaves the same way (I'm using x86_64 BTW).

Cheers,
    Olly