1 |
Brent Busby posted on Wed, 18 Feb 2015 22:19:37 -0600 as excerpted: |
2 |
|
3 |
> With all the arguing I see in the forums lately about libav versus |
4 |
> ffmpeg, I thought I'd ask an ffmpeg question of a different nature: |
5 |
> |
6 |
> Upstream, the ffmpeg people are releasing 2.5 as "stable." Some people |
7 |
> (who I'm not going to argue with, since I don't really have a position |
8 |
> on it myself) say the ffmpeg developers are reckless. However, I notice |
9 |
> that most other distros are at least shipping 2.2 (perhaps with lots of |
10 |
> distro patches to fix problems?). |
11 |
> |
12 |
> Gentoo's stable ffmpeg is currently 1.2.6, a fully maintained (i.e., not |
13 |
> defunct or deprecated) branch with ffmpeg upstream, apparently |
14 |
> maintained for really conservative users. |
15 |
> |
16 |
> All of this has me wondering...should I unmask? Both 2.2 and 2.5 are |
17 |
> available in Portage as ~amd64 ebuilds. Is anyone here doing this? Does |
18 |
> it unleash the hounds of hell (or the UNIX equivalent, lots of |
19 |
> unpleasant segfaults and codec problems)? I'm mostly wanting better |
20 |
> Bluray/WMV3/VC-1 support, but I don't know how that stands between the |
21 |
> various 1.2/2.2/2.5 branches. |
22 |
> |
23 |
> From talking to people who don't run Gentoo, I've found that it has a |
24 |
> reputation for being bleeding-edge, but at times, I've found to the |
25 |
> contraty that it can actually sometimes be overcautious, making it hard |
26 |
> to tell if you're missing out on things other distros are enjoying |
27 |
> because your upgrades are masked. Sometimes, package versions will stay |
28 |
> masked for months or years just because nobody got around to marking |
29 |
> them stable for amd64 and not because of any actual bugs. |
30 |
> |
31 |
> Anybody running ffmpeg 2.x without issues? If so, which one...2.2 or |
32 |
> 2.5? |
33 |
|
34 |
[Adding this note after I wrote the below. I think it ended up reading |
35 |
harsher than I intended. I know people choose stable for a reason and I |
36 |
really do respect that. Those people simply aren't me, nor that reason |
37 |
mine... and that certainly comes out in the below. But no personal, or |
38 |
even whole-group, offense intended. It's just not me. But it's your |
39 |
machine, not mine, and gentoo wouldn't be gentoo if it didn't respect |
40 |
your ultimate control of your machine, neither would I be me in the same |
41 |
case. So, umm... Read the below with that in mind. =:^) ] |
42 |
|
43 |
I think what you actually want to know (which isn't what you asked), is |
44 |
if anyone running _sta(b)le_ amd64, is running ffmpeg 2.x without issues. |
45 |
|
46 |
Because people running ~amd64, including me, if they're current (I am), |
47 |
are running ffmpeg-2.5.4. |
48 |
|
49 |
And in general, I've not seen issues. |
50 |
|
51 |
However, that's because the various other apps also in ~amd64 have in |
52 |
general already been updated to work with the newer ffmpeg. |
53 |
|
54 |
Which is the problem with stale^h^hble amd64, and sta(b)le gentoo in |
55 |
general. The newer ffmpeg can't be unmasked to sta(b)le until pretty |
56 |
much everything else in sta(b)le has been updated to work with it as well. |
57 |
|
58 |
And in fact, whenever there's something that has gotten as far behind on |
59 |
stable as ffmpeg has, in particular, for stuff like gcc, it's very likely |
60 |
the same general problem. On gentoo, for packages like gcc in |
61 |
particular, even ~amd64 (and ~arch in general) is generally quite behind, |
62 |
because before it's unmasked to ~arch, they want to ensure that all other |
63 |
packages (at the same ~arch) level have been patched to build with it. |
64 |
Which means it takes "forever" to unmask even to ~arch, because of all |
65 |
those other packages that have to be either updated first, or patched to |
66 |
build with the new gcc. Then when all that is done and it hits ~arch, |
67 |
the process starts over for sta(b)le; all those patched packages have to |
68 |
be sta(b)le keyworded before gcc itself can be sta(b)le keyworded. |
69 |
|
70 |
And yes, sometimes that does mean "stable" is "stale", no two ways about |
71 |
it. But it's a choice you make... |
72 |
|
73 |
But at least gentoo does make it real easy to package.keyword specific |
74 |
packages if you like, so you can be sta(b)le on most things while |
75 |
choosing to try arch or even masked versions if you dare. |
76 |
|
77 |
If you like you can do an equery depends ffmpeg, and see which packages |
78 |
that you actually have installed depend on it. You can then research |
79 |
each one to see if the version you currently have merged can function |
80 |
with the newer ffmpeg, and if not, you can of course package.keyword it |
81 |
~arch at the same time. |
82 |
|
83 |
With a bit of luck all the packages you have installed are already |
84 |
updated and it'll "just work" (perhaps with a rebuild of some of them) |
85 |
for you. If you're not that lucky, but still somewhat lucky, they've at |
86 |
least been updated with blockers or version ranges, so if you go to |
87 |
install the newer ffmpeg, portage will tell you which packages you have |
88 |
to keyword-unmask in ordered to proceed. |
89 |
|
90 |
Of course you also risk that they've not been updated yet, and simply |
91 |
quit working with no explanation, until you figure it out and unmask them |
92 |
so you get the newer version of them too. |
93 |
|
94 |
But at least with an equery depends ffmpeg, you can see how many packages |
95 |
you are risking, and decide whether it's worth the hassle from there, or |
96 |
if it's too much to risk/hassle and you prefer to wait for the full |
97 |
stable update after all. |
98 |
|
99 |
Of course another alternative is to "Dear Interwebs" the solution. You |
100 |
can post your equery results here and see if there's enough other people |
101 |
on sta(b)le but running a newer/package.keyworded ffmpeg with those |
102 |
packages as well, to compare notes. =:^) |
103 |
|
104 |
-- |
105 |
Duncan - List replies preferred. No HTML msgs. |
106 |
"Every nonfree program has a lord, a master -- |
107 |
and if you use the program, he is your master." Richard Stallman |