On 20.02.2013 13:19, Ivan Vučica wrote:
On Wed, Feb 20, 2013 at 10:24 AM, Fred Kiefer <[email protected]> wrote:
We don't add another timer here. The code just uses the same timer for a
different runloop mode. This should allow the existing invalidation code to
handle the timer correctly.
From your comment I understand that you don't know too much details about
our timer implementation. Let me state one point why I am reluctant to
commit this change without comments from Richard or anybody else who
understands that implementation: Some time ago we had a long standing bug
in base that resulted in timers that were added for two modes not being
handled correctly. There is a comment in the code now of this being fixed.
Even though the change in gui looks completely harmless, it could result
trigger a bug somewhere else.
Any downsides to having two timers, then, aside from uglier code? At least
for this release.
As I wrote the idea would be to reuse the same timer for both run loop
modes. This is definitely better than having two separate timers. The
specific problem I talk about was fixed in base a long time ago. I just
used that as an example that harmless changes in gui might lead to
unexpected issues.
If we do any change in that area, it should be the one I proposed. The
question is, should we do so before the release?
_______________________________________________
Gnustep-dev mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/gnustep-dev
