[FFmpeg-devel] Areas needing work

Derek Buitenhuis derek.buitenhuis at gmail.com
Mon Apr 6 20:24:35 CEST 2015


On 4/6/2015 4:43 AM, Michael Niedermayer wrote:
> * Post merge review&cleanup, changes merged from libav would benefit from
>   people reviewing and MUCH MUCH more importantly also people actually caring
>   about and taking responsibility not just dumping comments with the idea
>   "someone" will fix it.
>   Ideally each area would have a maintainer who takes care of it and
>   keeps it in good nick.

How would this actually work? Sit around and wait for merge commits and look at them?

I sometimes do, but most of the commits that are merged are pretty boring stuff, and
I tend not to read them.

> * fate clients
>   we lack a armcc test, a real android test
>   someone should take over haiku and sunos/solaris fate clients, my haiku VM just
>   keeps crashing

Is there a wiki somewhere of stuff that needs to be tested? I shut down all my FATE
instances 2 weeks ago for maintenance, but will be bringing them + more back up.

> * our fate server lacks the ability to display a user specified release
>   this makes the output look quite unwieldy
>   Timothy was working on a rewrite

[...]

> * Paid maintainers
>   a big problem is that we have areas in the codebase that have no maintainr or
>   rather noone with sufficient time+will+abilty.
>   It would help tremendiusly if maintainers could work full
>   time on maintaining code instead of just a few hours in their free
>   time

Where exactly would the money come from? Most, if not all of us have day jobs,
and/or are students.

> * ffserver lacks a regression test
>   code that is not tested will accumulate bugs. period
>   there are people who use ffserver, they clearly like the lack of regression
>   tests it seems as noone finds the need to look into changing that.

Drop ffserver. It is crap, and I have been saying this for years. It has no legitimate
use cases as far as I am concerned.

> * more people should help on https://trac.ffmpeg.org, like check that tickets
>   contain all information needed to reproduce and ask the reporter to provide
>   missing information, check if things are regressions, if they can be reproduced,
>   bisect regressions or ask the reporter to bisect and help him if needed
>   close tickets where there is not enough information to reproduce and the
>   reporter doesnt react for months. ATM carl does all that and more but
>   iam sure he wouldnt mind if more people would help him

I can only speak for myself and a couple others, but we stay well away from the bug
tracker because we cannot stand a certain prolific person there.

> * Stipends and Sponsorship programs
>   We do not have enough funds for basic things, for example OPW/Outreachy is
>   twice a year and to participate we have to pay at least one stipend
>   (about 6.5k USD). We participated in december 2014 and it allowed us to convert
>   all remaining libmpcodecs filters to the native system. resulted in multiple
>   new video filters and several block ciphers which are faster then all other
>   tested implementations. For may 2015 we only way after the last minute managed
>   to secure some funding so we are short of applicants now as for a long
>   time ffmpeg was not listed as a participating org.

And we have a bunch of massive failures too. I'm not 100% certain it's even worth it.
Not that we have been accepted in GSoC it seems like a better route, financially,
for attracting new people. Or, you know, people could contribute without cash
incentives - but I guess that concept is lost nowadays? :/

> * Decentralize
>   Too many things are done and admined by me, iam doing security updates to the
>   servers, doing the git merges, maintaining all releases, maintaining half the
>   fate clients, submitting data to coverity and doing all the crap that noone
>   else wants to do.

It's not really documented all the stuff which needs to be done, such as server updates,
or who handles them and pay, etc. Basically what Timothy said.

>   Does the community care so little that they want me to be a single point of
>   failure?

Nobody wants that. I don't recall any previous requests for help with the overhead work
such as server admin though.

>   We need people to take some of this over at least on the basis of a backup
>   in case iam unavailable for whatever reason, and i think it should ideally
>   be people who do not yet have their daily work shedule overflowing with ffmpeg
>   work but better new volunteers.

I don't mind pitching in a small amount, and I can do a bit during work hours perhaps
(whithin sanity/reason), but I also have a busy life with travel and significant
other, etc.

The main problem is it is not clear where or how to go about it, or what can be volunteered.

> * Conferences / exhibitions, public stuff
>   Theres a conference, linuxtag, fosdem, whatever, wherever in your local area
>   and you think FFmpeg should be represented there ?

FWIW, I attend FOSDEM and VDD representing both FFmpeg and Libav as well as $dayjob,
(just to make sure everyone on every side hates me ;)). I also talked at IBC at EBU's
booth about open source media, and will be in Berlin for SMPTE's circlejerk^Wforum
in May, as talking about open source media (FFmpeg especially). I'm not the only
one, and I think this is pretty well covered in recent years.

- Derek


More information about the ffmpeg-devel mailing list