[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

What to do with (packages like) Blender?

Hi folks.

It's been a while and I'm now really wondering what to do with
Blender. So that everyone can understand, I'm going to try and sum up
what I'm facing. Please note it's not intended to be a rant, rather a
summary of what I've to deal with.

 * Upstream doesn't really care about being distributed. The
   philosophy is rather "unzip and run". And by trying to distribute
   it, one might disable or even break some features (see below),
   which upstream doesn't like.

 * Similarly, when it comes to Release Candidates, the main idea is to
   get builds for every platform, meaning that between the first build
   and any other build, many patches get committed to fix various
   FTBFSes on Windows, Linux, and OS X (and even more?) platforms. No
   tag, people are supposed to use trunk when they see fit to give it
   a try. And what really matters is the availability of binaries.

 * Upstream doesn't really care about security. For example, there is
   a long standing patch to prevent using predictable filenames in
   /tmp (by moving the temporary directory to something like
   ~/.blender/tmp). Fortunately, other distributors seemed to care
   about sharing the patches and sdiscussing them.

 * Upstream uses a lot of embedded code copies. Most of them with
   patches. That means that one has to get rid of them, tweak the
   sources so as to be able to use system-wide libraries, and most
   importantly, deal with any breakages that may happen. That's my
   main problem: they have their own copy of ffmpeg, which is a huge,
   fast-moving library, and trying to ensure blender is usable at any
   time after ffmpeg updates isn't trivial (obvious API breakages are
   still OK, but subtle ones aren't). It looks like I was able to
   restore the broken sound output, but not the broken video output
   (not to mention it's been quite some time, like several
   weeks/months, and I'm quite feeling guilty about that). [I should
   note there's another upstream bugfix release (2.49a, Debian is at
   2.49), which I didn't investigate yet. Maybe that particular bug
   was noted by upstream and fixed, but I think my point stands

Which makes me wonder if there's a point in keeping blender as it
is. Maybe with someone able to dedicate a lot of time to it, that
might be feasible; but for one, I'm not really keen on spending lots
of time on problems upstream doesn't even care about (I've been kind
of flamed for posting a patch to add support for using system-wide

So: what should I do? I'm thinking about orphaning the package as a
first step, and if nobody takes care of it as much as it is needed,
just remove it from the distribution. Packaging would still be
available in a git repository, so if someone sometime picks it up,
it should be quite easy not to start from scratch again.

Thanks for any insights.


Attachment: signature.asc
Description: Digital signature

Reply to: