Re: GStreamer, PulseAudio and libvpx update

2016-02-03 Thread Andreas Enge
Hello,

On Wed, Feb 03, 2016 at 12:23:24PM +0800, 宋文武 wrote:
> In the 'gstreamer-update' branch we have following updates:
>   libvpx -> 1.5.0
>   pulseaudio -> 8.0
>   ao -> 1.2.0
>   gstreamer (and plugins) -> 1.6.3

I noticed you merged master into the branch. My opinion is that merging
instead of rebasing messes up the history and makes it rather unclear
what the differences in this branch are. So I would suggest the following:
Delete the branch, create a branch "wip-gstreamer" from security-updates
(not master!), try to build a few packages; after security-updates has been
applied to master, rebase wip-gstreamer and have it built by hydra.

Right now, the priority clearly is to finish security-updates, and we
cannot afford to build a second branch in parallel.

Are these four updates independent? Should they all be built together,
or should we do them one by one? Only 19 packages depend on ao, only 29 on
libvpx, so these could be done separately. Could they even go to master,
or do they depend on gstreamer (or pulseaudio) being updated first? 
pulseaudio has 162 dependent packages, so even these could maybe be built
separately (where by "separately" I mean in a different evaluation).

What do you think?

Andreas




Re: GStreamer, PulseAudio and libvpx update

2016-02-03 Thread Efraim Flashner
On Wed, 3 Feb 2016 11:25:22 +0100
Andreas Enge  wrote:

> Hello,
> 
> On Wed, Feb 03, 2016 at 12:23:24PM +0800, 宋文武 wrote:
>  [...]  
> 
> I noticed you merged master into the branch. My opinion is that merging
> instead of rebasing messes up the history and makes it rather unclear
> what the differences in this branch are. So I would suggest the following:
> Delete the branch, create a branch "wip-gstreamer" from security-updates
> (not master!), try to build a few packages; after security-updates has been
> applied to master, rebase wip-gstreamer and have it built by hydra.
> 
> Right now, the priority clearly is to finish security-updates, and we
> cannot afford to build a second branch in parallel.
> 
> Are these four updates independent? Should they all be built together,
> or should we do them one by one? Only 19 packages depend on ao, only 29 on
> libvpx, so these could be done separately. Could they even go to master,
> or do they depend on gstreamer (or pulseaudio) being updated first? 
> pulseaudio has 162 dependent packages, so even these could maybe be built
> separately (where by "separately" I mean in a different evaluation).
> 
> What do you think?
> 
> Andreas
> 

They should all be independant, and AFAIK even the gstreamer updates can be
independant of each other. The problem I ran into before was that
gst-plugins-good didn't compile correctly after updating some of the others.

That said, I think it would be worth it to try to pick off some of the
updates and see if they can be applied without breakage, so I'll see about
again while security-updates get built.

-- 
Efraim Flashner      אפרים פלשנר
GPG key = A28B F40C 3E55 1372 662D  14F7 41AA E7DC CA3D 8351
Confidentiality cannot be guaranteed on emails sent or received unencrypted


pgp0fx0FFZfow.pgp
Description: OpenPGP digital signature


Re: GStreamer, PulseAudio and libvpx update

2016-02-03 Thread 宋文武
Andreas Enge  writes:

> Hello,
>
> On Wed, Feb 03, 2016 at 12:23:24PM +0800, 宋文武 wrote:
>> In the 'gstreamer-update' branch we have following updates:
>>   libvpx -> 1.5.0
>>   pulseaudio -> 8.0
>>   ao -> 1.2.0
>>   gstreamer (and plugins) -> 1.6.3
>
> I noticed you merged master into the branch. My opinion is that merging
> instead of rebasing messes up the history and makes it rather unclear
> what the differences in this branch are. So I would suggest the following:
> Delete the branch, create a branch "wip-gstreamer" from security-updates
> (not master!), try to build a few packages; after security-updates has been
> applied to master, rebase wip-gstreamer and have it built by hydra.
Sure, merge does look weird.  Done as you said, thanks for the guide :)
>
> Right now, the priority clearly is to finish security-updates, and we
> cannot afford to build a second branch in parallel.
OK.
>
> Are these four updates independent? Should they all be built together,
> or should we do them one by one? Only 19 packages depend on ao, only 29 on
> libvpx, so these could be done separately. Could they even go to master,
> or do they depend on gstreamer (or pulseaudio) being updated first? 
> pulseaudio has 162 dependent packages, so even these could maybe be built
> separately (where by "separately" I mean in a different evaluation).
>
> What do you think?
Yes, they're independent.  I update gstreamer and add gst-plugins-bad,
the other are from Efraim Flashner's work.  Due to gst-plugins-good
failed to pass some tests with the update of libvpx and pulseaudio,
I think put them together may bring some luck.
well, it seem the test failure of gst-plugins-good is unreleated to
the update...




Re: GStreamer, PulseAudio and libvpx update

2016-02-03 Thread Efraim Flashner
On Wed, 03 Feb 2016 12:23:24 +0800
iyzs...@member.fsf.org (宋文武) wrote:

> In the 'gstreamer-update' branch we have following updates:
>   libvpx -> 1.5.0
>   pulseaudio -> 8.0
>   ao -> 1.2.0
>   gstreamer (and plugins) -> 1.6.3
> 
> This will lead to mess rebuilds.
> I have build gst-plugins-good, ffmpeg, mpv and pavucontrol,
> tested with pulseaudio-8.0, so I think it's safe to merge.
> Or should we start a job for the branch on hydra?
 
Depending on the number of packages to rebuild, giving hydra a head start of
a day or three is probably a good idea.
 
> Efraim Flashner  writes:
> 
>  [...]  
>  [...]  
>  [...]  
> With the gstreamer-update branch merged with current master, I don't
> meet any failure for my local x86_64 build of gst-plugins-good-1.6.3.
> 
> Do the failure still occur to you?

I just checked out origin/gstreamer-update again and gst-plugins-good failed
to build for me again.
FAIL: elements/splitmux
FAIL: elements/rtprtx

-- 
Efraim Flashner      אפרים פלשנר
GPG key = A28B F40C 3E55 1372 662D  14F7 41AA E7DC CA3D 8351
Confidentiality cannot be guaranteed on emails sent or received unencrypted


pgpWCwRlb3xpW.pgp
Description: OpenPGP digital signature


Re: GStreamer, PulseAudio and libvpx update

2016-02-03 Thread Andreas Enge
Hello,

On Wed, Feb 03, 2016 at 07:15:11PM +0800, 宋文武 wrote:
> Yes, they're independent.  I update gstreamer and add gst-plugins-bad,
> the other are from Efraim Flashner's work.  Due to gst-plugins-good
> failed to pass some tests with the update of libvpx and pulseaudio,
> I think put them together may bring some luck.
> well, it seem the test failure of gst-plugins-good is unreleated to
> the update...

thanks for getting back to this! I actually had forgotten to mention that
I would be willing to work on this, sorry.

So far, I tried the following: On x86_64, gst-plugins-good builds without
problem with the new version of ao that Efraim already pushed to master
and the new pulseaudio version. One test fails if I also apply the libvpx
update.

So it looks like we could also update pulseaudio in master; according to
"guix refresh -l" there are 168 dependent packages. Is this too much?

In any case, we should now wait for security-updates to be applied to
master, and probably keep the old version of libvpx until after the
gstreamer update.

Andreas




Re: GStreamer, PulseAudio and libvpx update

2016-02-03 Thread Efraim Flashner
On Wed, 3 Feb 2016 18:43:09 +0100
Andreas Enge  wrote:

> Hello,
> 
> On Wed, Feb 03, 2016 at 07:15:11PM +0800, 宋文武 wrote:
>  [...]  
> 
> thanks for getting back to this! I actually had forgotten to mention that
> I would be willing to work on this, sorry.
> 
> So far, I tried the following: On x86_64, gst-plugins-good builds without
> problem with the new version of ao that Efraim already pushed to master
> and the new pulseaudio version. One test fails if I also apply the libvpx
> update.
> 
> So it looks like we could also update pulseaudio in master; according to
> "guix refresh -l" there are 168 dependent packages. Is this too much?
> 
> In any case, we should now wait for security-updates to be applied to
> master, and probably keep the old version of libvpx until after the
> gstreamer update.
> 
> Andreas
> 

I tested the ao update which was obviously fine, but libvpx broke
gst-plugins-good. Its nice to know that pulseaudio doesn't break anything, I
assumed I was looking at a 2+ hour build session to really test it so I'm
glad I didn't have to be the one to do it in the end :) Debian managed to
upgrade libvpx to 1.5.0 with gstreamer at 1.6.3 so I'm a bit curious about
how they managed that.

-- 
Efraim Flashner      אפרים פלשנר
GPG key = A28B F40C 3E55 1372 662D  14F7 41AA E7DC CA3D 8351
Confidentiality cannot be guaranteed on emails sent or received unencrypted


pgpO91Zk73k9a.pgp
Description: OpenPGP digital signature


GStreamer, PulseAudio and libvpx update

2016-02-02 Thread 宋文武
In the 'gstreamer-update' branch we have following updates:
  libvpx -> 1.5.0
  pulseaudio -> 8.0
  ao -> 1.2.0
  gstreamer (and plugins) -> 1.6.3

This will lead to mess rebuilds.
I have build gst-plugins-good, ffmpeg, mpv and pavucontrol,
tested with pulseaudio-8.0, so I think it's safe to merge.
Or should we start a job for the branch on hydra?


Efraim Flashner  writes:

> On Thu, 21 Jan 2016 09:54:25 +0800
> iyzs...@member.fsf.org (宋文武) wrote:
> [...]
>> 
>> Sorry I didn't follow the issue.  What break on master due to the update
>> of libvpx?  Since hydra is busy with core-updates, I think we can work
>> on a separate branch, and pick it later after core-updates merged.
>
> Against my local gstreamer- and gst-plugin-*-1.7.1 I updated libvpx to 1.5.0
> and pulseaudio to 8.0 and gst-plugins-good-1.7.1 built with no issues. When I
> compiled gst-plugins-good-1.6.3 against libvpx-1.5.0 and pulseaudio-8.0 I got
> the same failure as before (which I forgot to write down), so I'm thinking
> this will mostly resolve itself when the gstreamer-1.8 release comes
> out.
With the gstreamer-update branch merged with current master, I don't
meet any failure for my local x86_64 build of gst-plugins-good-1.6.3.

Do the failure still occur to you?