Re: [xmlblaster-devel] More TestFailsafe

[Michele Laghi <laghi at swissinfo.org>]

Hi Martin

Martin Johnson wrote:
> Hi Michele
>
>
> > These callbacks are there to notify the client about state changes in
> > the client library. Some clients could ignore them, and some others
> > would like to take full control (for example by handling the queued
> > messages by themselves). The right place to do that stuff is in these
> > reached* methods.
>
>
> The semantics also provide the oppurtunity to catch reachedAlive no matter
> what default behaviour is (flush or not etc).
>
> The timeout on the sleeping thread is not going to be the only issue if
> network is
> also problematic (long timeouts etc) if lower level polices are  not in
> place.
>
> I guess what I'm trying to say is, could you have test on reachedAlive
> an brief sleep to continue instead of sleep with fingers crossed and then
> assert.
>
> If not just for the sake of demo.
Yes you'r right we could use the reachedAlive.
> > I am not able to reproduce that error. Could you have a closer look at it ?
>
>
> Yep.
>
>
> > I have no experience with that compiler but I am using gcc 2.95.3 and
> > Marcel gcc 3.2. Could you see which lexical_cast was failing ?
>
>
> XmlBlasterAccess.cpp line 34
>
> instanceName_ = lexical_cast<string>(this). This causes exception on my RH
> and gives memory location on win32 is this the intention?
Yes the intention was just to have a unique name. The memory location 
would fullfill this requirement. We were not able to reproduce the error 
but I guess I will take a unique Timestamp instead and the problem 
should be solved.
> > Are you trusting ddd ? I believe it is not handling threads right.
>
>
> Thanks for the tip. What do you use? just straight gdb? please advise I'm
> just a lowly win32 programmer who wishes for easy GUI.
ddd is just a GUI frontend to the gdb. If there are problems with ddd 
then you have trhe same problems with gdb. I am not really up to date on 
this issue. I heard however that there are pathches available meanwhile 
for gdb which should handle multithreading correctly.

Saluti
Michele

> Regards
> Martin