I'm working on a new version of this test module that implements a
one-thread-one-interpreter model as George Petasis pointed out. It seems
to work but there a few problems still to be solved. I hope I can commit
next week a version that should be able to parse / execute scripts
without initialization handling, just as a testbed for doing a tests and
prove the robustness of the model. This turned out to be a critical task
as the previous module started tp crash when the server parameters were
set to values that emulated extreme resource depletion.
-- Massimo
On 19-07-2014 22:52, Massimo Manghi wrote:
The problem seems to be 8.6 specific whereas the module works with
Tcl 8.5 without the mutex protecting the call to Tcl_EvalObjEx. Thus
it looks like 8.6 doesn't like to work with threads when the
environment embedding Tcl is multithreaded.
I get segfaults with random backtraces stemming from Tcl_EvalObjEx
which suggests threads are spoiling unprotected data all around in
several Tcl functions. NRE functions always appear in the bt output,
but this is the engine for Tcl 8.6 so it's natural to find them.
If anyone with a closer understanding of the Tcl core (Harald?,
Arnulf?) has suggestions I would appreciate if they could express an
opinion, I think I'm going to bring up the issue anyway on TCLCORE,
just wanted to be prepared at the best to explain what's going on
-- Massimo
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
