@ Holtz
That is a known fact in software development. Maintaining one designated software version with bug fixes and potentially back ports, while at the same time proceed with development is a strain on the recourses.
In my ignorance I thought that, since bugs are fixed any way, it could be as simple as committing the fixes to two branches, ‘trunk’ as we have already and another one where no new features, only applicable bug fixes would be admitted. I guess that what you are saying is that fixes for the same bug aren’t the same for an evolving branch like ‘trunk’ than they would be for a ‘no new feature’ branch ?
the possibility to compile from the sources where the bugs are fixed.
I watch what’s committed daily and build almost every day too. Although the bug fixes that I wish to see in a ever more stable official version are in those daily builds, more bug are created by committing new code. Stability is never insured, as it is only reasonable to expect. I’ve had to keep two SVN three, building them alternatively, in case one build just won’t succeed. I made a small script to save files and re-open them right away in case the file becomes unreadable. I have autosave saving every minute two minutes so it is easy to run Blender in debug mode with the file that caused the problem. Right now Campbell, Sergey, Brecht and a few others like Bastien, Nicholas… seem to keep a close watch on the bug tracker and are even fixi stuff that they notice by themselves. We’ll see: after all they also have other duties.
Redistributing resources available to maintain a long term release
I wasn’t thinking of something as involved as an LTS: just fixes for bugs that are done anyway and no new features.
If you yourself are willing to invest in one of these three options, go ahead and buckle down.
I am learning to code: slow process. For the moment my (maybe too lofty) goal is to create an environment for artisan furniture and renovation shops where the metaphor would be closer to the practice of the trade. One thing for sure: I’ll fix my own bugs.
@ Sanctuary
and do not include experimental changes or features like the builds you can find graphicall
Still, the code that gets committed is usually not thoroughly tested and hence more buggy than that of any official release.
On the bug tracker the devs have been working really hard, the amount of bugs they fixed daily is impressive.
Simply breathtaking.
But still, if you can , try to report the problems you experienced and give the devs the most hints possible on reproducing it.
You can read in my response to Holtz all I do to report bugs.
@temujin143
I think it is fair to complain about stability after Blender 2.65 but not while we are in 2.63 and 2.64.
I think that you are right on that. Nevertheless, asking around if there doesn’t exist one fairly simple way to enhance the final stability of each version != complaining.
@ ParticleXY
What happens to these builds? I see they are all revisions of 2.63 but will code from these eventually make it into a 2.63a release?
AFAIK they are builds of the trunk not pure revisions of the last official release, which is exactly what I’d like to see and very subject of this thread. The include new code destined to the next version of Blender with its own lot of bugs hence the ‘less stable’ warning.
Not trying to complain, just want to figure out how the release philosophy works
So am I, exactly: making my education.
I really commend the efforts of the developers in squashing so many bugs, but it seems that sometimes their efforts are offset by only being available in (possibly buggy) new revisions and not official bugfix releases.
Wise words IMHO.
Fair enough, but there is also an influx of downloads when the official version is released, so a certain amount of bugs only get found a short while after release. Relying only on release candidates as a bug benchmark ignores the nature of the general downloading public, does it not?
Your observation is again quite right: Blender get its most thorough testing from the day of the official release on. The common response to that is that is the responsibility of the users to test the RCs. This is unfair to many. Firstly I have a feeling that Blender has more quality users with each passing day who are responsible for more bug reports than ever. Secondly it is not everyone who can afford the risk that RCs present while racing to meet the deadlines.
@abc123
I would say that most of the buildbot builds are actually more stable than the “stable” blender, featuring bug fixes for problems that were in the “stable” release.
That’s been my experience too. The problem is that when new features or even new code breaks some part of Blender it is usually in some entirely unexpected way and sometimes in apparently totally unrelated areas, parts that may have always worked just fine foe you. This is hard to anticipate, can be costly, although it is risk that can be minimized by taking a number of precautions.
@san
I understand the need for a more conservative bugfix-only “LTS” release, but even though it seems like Blender has a ton of developers with its current pace, there’s no one who has the time/motivation to maintain such a release.
Again, I wasn’t thinking of an LTS: just applying the fixes to the official version until the next one gets released.