Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [cdt-dev] DSF timeouts

 1. There's always some timeout for every single command sent to GDB.

I think this kind of timeout could be implemented in AbstractMIControl.  A periodic runnable could check the time stamps of outstanding commands and complete them after a set timeout.

There is a more general problem though, in that programming errors and runtime exception can cause requests to never get completed.  I think Vladimir you may be referring to this problem wrt to the timeouts.  This has been a chronic problem with DSF since the start.  To solve it we could make up a request monitor registry that tracks requests as they're passed through the call hierarchy and automatically completes them when they get "lost".

Cheers,
Pawel


On 06/29/2011 08:20 AM, Vladimir Prus wrote:
On Monday, June 27, 2011 18:27:21 Marc Khouzam wrote:
-----Original Message-----
From: cdt-dev-bounces@xxxxxxxxxxx
[mailto:cdt-dev-bounces@xxxxxxxxxxx] On Behalf Of Vladimir Prus
Sent: Friday, June 24, 2011 1:50 PM
To: cdt-dev@xxxxxxxxxxx
Subject: [cdt-dev] DSF timeouts


Hello,

some time ago I've asked whether DSF has, in general, timeouts for all

GDB commands that it sends:
	http://dev.eclipse.org/mhonarc/lists/cdt-dev/msg19523.html

I was pointed to a specific Bugzilla issue about 'interrupt' command:
	https://bugs.eclipse.org/bugs/show_bug.cgi?id=310274

and that specific issue was fixed, but I'm having trouble
understanding
whether, in current DSF HEAD:

1. There's always some timeout for every single command sent to GDB.
No, we don't have that.  It has been suggested before and I think it
would add a another level of safety, but it has not proven urgent
enough for anyone to contribut it.
Do you have a suggestion where such logic can be added?

2. There's a timeout for the launch sequence as a whole.
There is a 30 second timeout for GDB to be started.
However, the rest of the launch sequence does not have a timeout.
I don't think it would be very hard to add, probably in GdbLaunchDelegate.
Again, could you suggest the best way? It seems that ideally, the 'execute'
method of the executor would have a timeout parameter.

- Volodya




Back to the top