RE: [GD-Windows] Mutexes and win2k

[Martin S. <ms...@cl...>]

Well wadda know thats embarassing, thers a clever server that does it for us
and this allegedly is not ours (everything in my outbox is in text only) , I
will try and get around, till then all you unix guys and elm users will have
to wait.

Martin

(and see if this works now)

-----Original Message-----  From: Martin Slater
[mailto:ms...@cl...]
Sent: 28 November 2001 17:44
To: 'gam...@li...'
Subject: RE: [GD-Windows] Mutexes and win2k




As Rich (no not on this list to date but from DX) was kind enough to point
out i was negligent in sending in html (didn't realise it had, but hey ho
outlook has been duly instructed)so here is again for all you unix junkies
out there, roll it............
Morning all, 
Wondering if anyone has come across any problems with mutexes uner win2k, i
am creating a named mutex from one process and opening it up from another,
if the app which creates the mutex is run from the debugger (or win98) the
client can open the mutex happily but if the server is run normally then
9/10 times the client refuses to open the mutex (returning file not found) ,
but just occaisonally it will open it ok.  I'd swear this worked ok in the
past but at some point broke unbeknown to me as it was either running on a
98 box or being debugged on the 2k box. A simple test program works fine
though.  
Something else i noted which may or may not be relevent (all this is using
vc6+sp5) was the following code 
        string tmp_str; 
        tmp_str = server_name; 
        tmp_str += access_mutex_name; 
was not working as expected when being executed in a thread, yielding
tmp_str = access_mutex_name (ignoring the server name), but when run from
the main thread it got it right, anyone know if this is some problem with
the dinkumware stl?
Everything is being linked against the multithreaded dll version of the rtl.

Got me stumped 
Any ideas? 
cheers 
Martin 



PS> Thinking aloud, could it be something to do with something not being
correctly labeled volatile  and the debugger is causing the registers to be
reloaded anyway somehow? still doesn't explain why it works fine on 98
tho.hmmm