🚀 go-pugleaf

-- 
Andreas Gehring							email: andrew@medap3.cs.tu-berlin.de

I post this article for a friend. you can send the answer to my email address -
I will give it to him:

basic situation:
In my client server architecture every clientprocess logging to the serverprocess gets a subserverprocess specifically linked to that client via two named pipes, each of them readable at the ownerside. So, the client makes a read-pipe, waits for connection by the subserver and furtheron for requested data and vice versa the subserver on its side. A second and third client logging to the server will get their own subservers together with their own named pipes. The pipenames are generated through the client'


s processid plus an extention, so all pipenames are unique.

problem:
Having simply connected one client to the subserver architecture prooves working fine, logging a second client parallel to the first to the server ends in a termination of the first client. This happens probably due to the access to the first clients's pipe, because both client processes generate the same pipehandles (numbers starting with 3,4,...). Resumed, the first client fails in writing its request into the subserverpipe (ERROR_PIPE_BROKEN), and the second subserver fails in reading the second client'


s request (ERROR_DISCARDED).

If anybody has a solution to my problems - please let me know!