WebRTC: Bot for autorolling WebRTC+libjingle in Chromium DEPS.

WebRTC: Bot for autorolling WebRTC+libjingle in Chromium DEPS.

This depends on http://crbug.com/539527.

BUG=666726

[kjellander_chromium, 2015-10-05 19:35:47]

kjellander@chromium.org changed reviewers:
+ machenbach@chromium.org

[kjellander_chromium, 2015-10-05 19:35:47]

wow this was quickly done by copy-pasting your recipe :)

https://codereview.chromium.org/1384273003/diff/1/scripts/slave/recipes/webrt...
File scripts/slave/recipes/webrtc/auto_roll_chromium_deps.py (left):

https://codereview.chromium.org/1384273003/diff/1/scripts/slave/recipes/webrt...
scripts/slave/recipes/webrtc/auto_roll_chromium_deps.py:29:
api.gclient.apply_config('v8_bleeding_edge_git')
Our script at
https://code.google.com/p/chromium/codesearch#chromium/src/tools/roll_webrtc.py
ensures the directories to run roll-dep-svn in are up to date using 'git fetch
origin'.

[Michael Achenbach, 2015-10-06 09:31:55]

lgtm - not sure if it'll work - but green light for trying it out :)

https://codereview.chromium.org/1384273003/diff/1/scripts/slave/recipes/webrt...
File scripts/slave/recipes/webrtc/auto_roll_chromium_deps.py (right):

https://codereview.chromium.org/1384273003/diff/1/scripts/slave/recipes/webrt...
scripts/slave/recipes/webrtc/auto_roll_chromium_deps.py:48: 'owner':
'webrtc-autoroll@chromium.org',
You have an account now?

https://codereview.chromium.org/1384273003/diff/1/scripts/slave/recipes/webrt...
scripts/slave/recipes/webrtc/auto_roll_chromium_deps.py:68: 'import time;
time.sleep(20)',
Not sure if you need that, but it doesn't hurt much. In our case without waiting
we can run into:

1. CQ landing a roll (and closing the issue).
2. gnumbd not processing it yet
3. Auto roller checking for active roll and trying to roll the same again.
4. gnumbd processes the first roll finally.
5. The new roll creates patch failures on the CQ

But even with waiting, this can happen in a gnumbd outage.

https://codereview.chromium.org/1384273003/diff/1/scripts/slave/recipes/webrt...
scripts/slave/recipes/webrtc/auto_roll_chromium_deps.py:77:
cwd=api.path['checkout'],
Sure this works? This is from a chromium checkout.

[kjellander_chromium, 2015-10-06 09:58:10]

https://codereview.chromium.org/1384273003/diff/1/scripts/slave/recipes/webrt...
File scripts/slave/recipes/webrtc/auto_roll_chromium_deps.py (right):

https://codereview.chromium.org/1384273003/diff/1/scripts/slave/recipes/webrt...
scripts/slave/recipes/webrtc/auto_roll_chromium_deps.py:48: 'owner':
'webrtc-autoroll@chromium.org',
On 2015/10/06 09:31:55, Michael Achenbach wrote:
> You have an account now?

No, not yet. I might try this with my own account while waiting for that one.

https://codereview.chromium.org/1384273003/diff/1/scripts/slave/recipes/webrt...
scripts/slave/recipes/webrtc/auto_roll_chromium_deps.py:68: 'import time;
time.sleep(20)',
On 2015/10/06 09:31:55, Michael Achenbach wrote:
> Not sure if you need that, but it doesn't hurt much. In our case without
waiting
> we can run into:
> 
> 1. CQ landing a roll (and closing the issue).
> 2. gnumbd not processing it yet
> 3. Auto roller checking for active roll and trying to roll the same again.
> 4. gnumbd processes the first roll finally.
> 5. The new roll creates patch failures on the CQ
> 
> But even with waiting, this can happen in a gnumbd outage.

As you say it doesn't hurt so I'll keep it for initial tests.

https://codereview.chromium.org/1384273003/diff/1/scripts/slave/recipes/webrt...
scripts/slave/recipes/webrtc/auto_roll_chromium_deps.py:77:
cwd=api.path['checkout'],
On 2015/10/06 09:31:55, Michael Achenbach wrote:
> Sure this works? This is from a chromium checkout.

Yeah, we moved the script from WebRTC into Chromium a while back:
https://code.google.com/p/chromium/codesearch#chromium/src/tools/roll_webrtc.py

I'm not entirely sure about the --wait-for-trybots thing since I don't think
anyone is using it but I wrote it with a bot like this in mind a while back, so
I'll give it a try. With that the build would go on until the CL is landed
instead of checking back over and over (resulting in many no-op builds, which
makes the overview harder to read).

[kjellander_chromium, 2015-10-26 08:40:29]

PTAL since I added LKGR lookup step now.
This should not be landed until https://codereview.chromium.org/1416773007/ is
in though.

https://codereview.chromium.org/1384273003/diff/1/scripts/slave/recipes/webrt...
File scripts/slave/recipes/webrtc/auto_roll_chromium_deps.py (right):

https://codereview.chromium.org/1384273003/diff/1/scripts/slave/recipes/webrt...
scripts/slave/recipes/webrtc/auto_roll_chromium_deps.py:48: 'owner':
'webrtc-autoroll@chromium.org',
On 2015/10/06 09:58:09, kjellander (chromium) wrote:
> On 2015/10/06 09:31:55, Michael Achenbach wrote:
> > You have an account now?
> 
> No, not yet. I might try this with my own account while waiting for that one.

Now I have an account (crbug.com/539527).

https://codereview.chromium.org/1384273003/diff/1/scripts/slave/recipes/webrt...
scripts/slave/recipes/webrtc/auto_roll_chromium_deps.py:77:
cwd=api.path['checkout'],
On 2015/10/06 09:58:09, kjellander (chromium) wrote:
> On 2015/10/06 09:31:55, Michael Achenbach wrote:
> > Sure this works? This is from a chromium checkout.
> 
> Yeah, we moved the script from WebRTC into Chromium a while back:
>
https://code.google.com/p/chromium/codesearch#chromium/src/tools/roll_webrtc.py
> 
> I'm not entirely sure about the --wait-for-trybots thing since I don't think
> anyone is using it but I wrote it with a bot like this in mind a while back,
so
> I'll give it a try. With that the build would go on until the CL is landed
> instead of checking back over and over (resulting in many no-op builds, which
> makes the overview harder to read).

It's worth mentioning that this will wait for max 1 hour, and that will output
to stdio every 30 seconds so the build step shouldn't die. I might take this
away if it doesn't work but I like the idea about having less no-op builds that
do nothing, which is why I wanted to at least give it a try.

[Michael Achenbach, 2015-10-26 08:54:38]

lgtm

https://codereview.chromium.org/1384273003/diff/1/scripts/slave/recipes/webrt...
File scripts/slave/recipes/webrtc/auto_roll_chromium_deps.py (right):

https://codereview.chromium.org/1384273003/diff/1/scripts/slave/recipes/webrt...
scripts/slave/recipes/webrtc/auto_roll_chromium_deps.py:48: 'owner':
'webrtc-autoroll@chromium.org',
On 2015/10/26 08:40:29, kjellander (chromium) wrote:
> On 2015/10/06 09:58:09, kjellander (chromium) wrote:
> > On 2015/10/06 09:31:55, Michael Achenbach wrote:
> > > You have an account now?
> > 
> > No, not yet. I might try this with my own account while waiting for that
one.
> 
> Now I have an account (crbug.com/539527).

Acknowledged. FYI: You might need to log on to the bot once and log in to
rietveld with this account so that credentials are stored.

You might also want to add a git config for that account on the bot. Otherwise
your CL author might end up being called chrome-bot.

https://codereview.chromium.org/1384273003/diff/1/scripts/slave/recipes/webrt...
scripts/slave/recipes/webrtc/auto_roll_chromium_deps.py:68: 'import time;
time.sleep(20)',
On 2015/10/06 09:58:09, kjellander (chromium) wrote:
> On 2015/10/06 09:31:55, Michael Achenbach wrote:
> > Not sure if you need that, but it doesn't hurt much. In our case without
> waiting
> > we can run into:
> > 
> > 1. CQ landing a roll (and closing the issue).
> > 2. gnumbd not processing it yet
> > 3. Auto roller checking for active roll and trying to roll the same again.
> > 4. gnumbd processes the first roll finally.
> > 5. The new roll creates patch failures on the CQ
> > 
> > But even with waiting, this can happen in a gnumbd outage.
> 
> As you say it doesn't hurt so I'll keep it for initial tests.

Actually you probably do need it. Also here a race can happen when CQ closes the
issue and this script passes the open-issue check, but gnumbd is slow and does
not yet stamp the last roll.

In case of gnumbd outages or occasional hiccups you might still get the problem.
The result would be a duped roll attempt that causes patch failures on the
trybots. Easy to fix by just closing the CL, but still annoying.

I wanted to work on a real fix at some point, which would be checking the
pending/heads/master branch for the currently rolled version.

https://codereview.chromium.org/1384273003/diff/1/scripts/slave/recipes/webrt...
scripts/slave/recipes/webrtc/auto_roll_chromium_deps.py:77:
cwd=api.path['checkout'],
On 2015/10/26 08:40:29, kjellander (chromium) wrote:
> On 2015/10/06 09:58:09, kjellander (chromium) wrote:
> > On 2015/10/06 09:31:55, Michael Achenbach wrote:
> > > Sure this works? This is from a chromium checkout.
> > 
> > Yeah, we moved the script from WebRTC into Chromium a while back:
> >
>
https://code.google.com/p/chromium/codesearch#chromium/src/tools/roll_webrtc.py
> > 
> > I'm not entirely sure about the --wait-for-trybots thing since I don't think
> > anyone is using it but I wrote it with a bot like this in mind a while back,
> so
> > I'll give it a try. With that the build would go on until the CL is landed
> > instead of checking back over and over (resulting in many no-op builds,
which
> > makes the overview harder to read).
> 
> It's worth mentioning that this will wait for max 1 hour, and that will output
> to stdio every 30 seconds so the build step shouldn't die. I might take this
> away if it doesn't work but I like the idea about having less no-op builds
that
> do nothing, which is why I wanted to at least give it a try.

Acknowledged.

[kjellander_chromium, 2015-10-26 08:57:20]

The CQ bit was checked by kjellander@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2015-10-26 08:57:32]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1384273003/20001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1384273003/20001

[commit-bot: I haz the power, 2015-10-26 09:00:16]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2015-10-26 09:00:17]

Dry run: This issue passed the CQ dry run.

[kjellander_chromium, 2016-11-18 13:59:16]

Description was changed from

==========
WebRTC: Bot for autorolling WebRTC+libjingle in Chromium DEPS.

This depends on http://crbug.com/539527.

BUG=433305
==========

to

==========
WebRTC: Bot for autorolling WebRTC+libjingle in Chromium DEPS.

This depends on http://crbug.com/539527.

BUG=666726
==========