[Vortex] More strange behavior during MSG/ANS pattern

Francis Brosnan Blazquez francis at aspl.es
Thu Oct 18 18:59:53 CEST 2007


Hi Chris,

It is confirmed. There is a bug inside the vortex engine causing to
produce fault frames while using the ANS/NUL pattern. It is weird
because regression tests are properly executed (without valgrind
exection). Even the following test [1] and [2] and properly executed
but, executed from different hosts (not localhost), produces the wrong
frame.

I would recomend to use a MSG/RPY pattern until a fix is proposed. The
ANS/NUL emulation is pretty forward:

1. Client app issue a MSG frame requesting the file.

2. Server app, replies (RPY) to the MSG received accepting to transfer
the file.

3. Server app, just after sending the RPY, starts issuing one MSG frame
for each corresponding ANS that would be generated, issuing a final
empty MSG emulating the NUL frame.

4. Client app, replies to all MSG received with an empty RPY, accepting
the content received. In the other hand, the server can just discard all
RPY messages received (they are just acknoledges).

5. Client app, in a final step detects the empty MSG received as a proof
that the file was completed transferred.

The point with MSG/RPY is that all MSG must have its associated RPY.

Cheers!

[1]
https://dolphin.aspl.es/svn/publico/af-arch/trunk/libvortex/test/vortex-file-transfer-server.c

[2]
https://dolphin.aspl.es/svn/publico/af-arch/trunk/libvortex/test/vortex-file-transfer-client.c

> Attached is the log from the server side. I agree with the memory
> management prognosis because of the behavior I am witnessing. 
> 
> If it helps, here's my tools/environment:
> Visual Studio 2005
> MFC & C++ 
> Windows XP SP2 and Windows Vista Pro

-- 
Francis Brosnan Blazquez <francis at aspl.es>
Advanced Software Production Line, S.L.




More information about the Vortex mailing list