Am 01.02.2012 21:31, schrieb Clemens:
On Thu, 02 Feb 2012 00:09:17 +1300, Stephan Aßmus <superstippi@xxxxxx> wrote:Am 01.02.2012 11:34, schrieb Clemens:On Thu, 26 Jan 2012 06:10:41 +1300, Axel Dörfler <axeld@xxxxxxxxxxxxxxxx> wrote:+ BAutolock _(sBitmapListLock); + sBitmapList.RemoveItem(this);I haven't read the actual code, but does this also work okay with no-server-link bitmaps?haven't checked it, also overlay bitmaps are not handled yet. They probably have to be handled separately on a server restart...If this is not properly tested, why not put it into a feature branch? The app_server is a critical component that already worked pretty stable at one time. As can be witnessed by the mail daemon changes, committing unfinished and not enough tested changes to critical components is pretty damaging and does not necessarily lead to other developers jumping in to bring everything back to stable operation. Even if they do they might not have enough time and meanwhile users are affected. It appears we have too few resources for that kind of workflow.Since the app_server should not crash this code should never be reached by a normal user, however, if it crashes there is a better changes now that stuff can be rescued.
Maybe I am missing something, but if you had to redesign some existing code as ground work for the restart feature to be possible, then that redesigned code is in use without the restart feature being executed at all. My mail was specifically in response to your comment about not having tested the bitmap changes in all situations.
I'm very sceptical that working on a branch would lead to any improvement. In this special case I just worked on it because I was curious if its easily possible. By committing it to a branch on github the changes was very high that neither I nor somebody else ever would have looked at it again. Now I got some feedback and I will work on it / may look into the bitmap stuff again :)
Again, I may be remembering something wrong, but I thought that if you create a branch in the official repository, then the commit notification channels do work. I fully agree about branches hosted elsewhere to have too little or no visibility. In any case, you got feedback since the commit mail was seen, not because you commited to the master branch, at least I would like to think so. (My own feedback excluded. ;-)
Best regards, -Stephan