Assembla home | Assembla project page
 

Ticket #7 (new enhancement)

Opened 4 months ago

Last modified 4 months ago

make hg operations cancellable (was: reduce number of timeouts)

Reported by: chyssler Assigned to:
Priority: normal Milestone: 1.2
Component: mercurialeclipse Version:
Severity: normal Keywords:
Cc:

Description

I'm quite sure that having individual timeouts for each command is unnecessary They could probably be grouped into long running and short running tasks then users would only need to enter two values for these.

Attachments

Change History

08/01/08 16:20:49 changed by bastiand

Hm. I like to have the choice. Having a longer timeout on clone than on pull is an example where the simplified approach might fail. Although you could as well have three timeout settings (low/medium/high).

08/04/08 07:30:06 changed by chyssler

The usage of the timeouts are either

  1. as a guard against locking up the ui
  2. as a means of supporting artifical cancel of too long running operations

In the best of worlds, a) should not be needed and b) should be provided by a cancel button in the progress manager. I wonder if such a cancel can be safely implemented? What happens to the repository if we terminate the process doing a commit/pull/push? I'd like to think hg is transactional, and I belive it it. So a cancel should be possible to implement.

08/10/08 12:34:57 changed by bastiand

  • summary changed from reduce number of timeouts to make hg operations cancellable (was: reduce number of timeouts).

hg is transactional. There are only very few operations that are not transactional, like the current qrebase SoC implementation.

So cancelling should be the way to go. Changing subject...

08/12/08 21:16:51 changed by chyssler

Now there seems to be no efficent way of polling a Process if its done or not (mep used to poll, but that is inefficent).

This implies we need to launch the process in a separate thread, poll our thread for completion or terminate if user wants to cancel.

Abit complex

08/17/08 11:26:51 changed by chyssler

Its not that bad if one uses the new java.util.concurrent facilities. I've a patch that prepares AbstractShellCommand? to be cancelable.


Add/Change #7 (make hg operations cancellable (was: reduce number of timeouts))