Re: More multi-threading questions...

[Justin Schoeman <justin expertron co za>]

*BLUSH*

Rule #1: when copying your code into a separate tree for testing, 
remember to change the base directory in your makefile... I was still 
using the include files from my main tree, but modifying the ones in the 
test tree...

I should have been suspicious when everything always compiled cleanly 
after making modifications ;-)

Anyway, now that I am _actually_ using g_main_loop_quit() to shut down 
the main thread, I can now shut down the server cleanly!

The occasional assertion - aborts on startup are still there though, but 
I have found no way to replicate them in gdb - they seem to be timing 
related, and gdb affects the timing too much...

As to valgrind, for some reason, when I run the server under valgrind, 
it does not correctly pass strings in client corba requests, so when it 
does a name service bind, it ends up binding with an empty string for a 
name... No idea why this happens. I will try memprof next.

Thanks again,
-justin

Michael Meeks wrote:
> Hi Justin,
> 
> On Tue, 2003-09-09 at 15:37, Justin Schoeman wrote:
> 
>>The thing is, no shutdown functions have been called yet. Only 
>>g_main_loop_quit() has been called, no CORBA shutdown functions...
> 
> 
> 	How then can that file descriptor possibly be -1 ? it's only set to -1
> in the giop_shutdown function; what happens if you put a break-point in
> giop_shutdown to see what's calling that.
> 
> 
>>Hmmm... All my apps bomb out completely in valgrind - CORBA parameters 
>>seem to be corrupted somehow.  I will try memprof next.
> 
> 
> 	You need to compile ORBit2 with --enable-purify to get rid of the
> uninitialized memory access type warnings.
> 
> 	Regards,
> 
> 		Michael.
>