Reland of predictors: Basic browsertest checks predictor learning.

Reland of predictors: Basic browsertest checks predictor learning.

Reason for reland:
The reason of the test flakiness was fixed.

ResourcePrefetchPredictor needs to be initialized before learn any
data. Initialization starts when the first main frame load complete for
given profile. In our case it was about:blank page after tab creation.
This initialization requires work in DB thread to be done. So there was
a race condition in which next navigation could appear earlier than
initialization in DB thread will be done.
This bug could be easily reproduced by adding sleep in
ResourcePrefetchPredictorTables::GetAllData function.

For now we explicitly check that initialization of
ResourcePrefetchPredictor was done before any navigation.

Original issue's description:
> Revert of predictors: Basic browsertest checks predictor learning. (patchset #9 id:160001 of https://codereview.chromium.org/2449323005/ )
>
> Reason for revert:
> Created a flaky test: first occurrence at
>
> https://build.chromium.org/p/chromium.chromiumos/builders/Linux%20ChromiumOS%20Tests%20%281%29/builds/28933
>
> Original issue's description:
> > predictors: Basic browsertest checks predictor learning.
> >
> > First stage for implementing ResourcePrefetchPredictor browsertests is to check
> > that the prefetch database is really filled after some navigations in browser.
> > Browsertest class provides convenient way to declare expected resources in the
> > test and then checks expectations internally that allows to write readable
> > tests.
> >
> > BUG=650253
> >
> > Committed: https://crrev.com/e636e3bcc172f220dfadfd2121d238c9908185bb
> > Cr-Commit-Position: refs/heads/master@{#429966}
>
> TBR=pasko@chromium.org,lizeb@chromium.org,alexilin@chromium.org
> # Skipping CQ checks because original CL landed less than 1 days ago.
> NOPRESUBMIT=true
> NOTREECHECKS=true
> NOTRY=true
> BUG=650253
>
> Committed: https://crrev.com/3250e71949fa7d226e9c843ee28277b988306229
> Cr-Commit-Position: refs/heads/master@{#430137}

BUG=650253
Committed: https://crrev.com/f2d707775a8184e7bd18d2334ff2a679f3a90439
Cr-Commit-Position: refs/heads/master@{#430561}

[alexilin, 2016-11-07 09:52:56]

Created Reland of predictors: Basic browsertest checks predictor learning.

[pasko, 2016-11-07 13:36:04]

I assume this should wait until we have a better idea of the reasons of the
flake.

[alexilin, 2016-11-07 17:03:13]

Description was changed from

==========
Reland of predictors: Basic browsertest checks predictor learning. (patchset #1
id:1 of https://codereview.chromium.org/2470143007/ )

Reason for revert:
Reland

Original issue's description:
> Revert of predictors: Basic browsertest checks predictor learning. (patchset
#9 id:160001 of https://codereview.chromium.org/2449323005/ )
> 
> Reason for revert:
> Created a flaky test: first occurrence at
> 
>
https://build.chromium.org/p/chromium.chromiumos/builders/Linux%20ChromiumOS%...
> 
> Original issue's description:
> > predictors: Basic browsertest checks predictor learning.
> >
> > First stage for implementing ResourcePrefetchPredictor browsertests is to
check
> > that the prefetch database is really filled after some navigations in
browser.
> > Browsertest class provides convenient way to declare expected resources in
the
> > test and then checks expectations internally that allows to write readable
> > tests.
> >
> > BUG=650253
> >
> > Committed: https://crrev.com/e636e3bcc172f220dfadfd2121d238c9908185bb
> > Cr-Commit-Position: refs/heads/master@{#429966}
> 
> TBR=pasko@chromium.org,lizeb@chromium.org,alexilin@chromium.org
> # Skipping CQ checks because original CL landed less than 1 days ago.
> NOPRESUBMIT=true
> NOTREECHECKS=true
> NOTRY=true
> BUG=650253
> 
> Committed: https://crrev.com/3250e71949fa7d226e9c843ee28277b988306229
> Cr-Commit-Position: refs/heads/master@{#430137}

TBR=pasko@chromium.org,lizeb@chromium.org,mathp@chromium.org
# Not skipping CQ checks because original CL landed more than 1 days ago.
BUG=650253
==========

to

==========
Reland of predictors: Basic browsertest checks predictor learning.

Reason for revert:
The reason of the test flakiness was fixed.

ResourcePrefetchPredictor needs to be initialized before learn any data.
Initialization starts when the first main frame load complete for given profile.
In our case it was about:blank page after tab creation. 
This initialization requires work in DB thread to be done. So there was a race
condition in which next navigation could appear earlier than initialization in
DB thread will be done.

For now we explicitly check that initialization of ResourcePrefetchPredictor was
done before any navigation.

Original issue's description:
> Revert of predictors: Basic browsertest checks predictor learning. (patchset
#9 id:160001 of https://codereview.chromium.org/2449323005/ )
> 
> Reason for revert:
> Created a flaky test: first occurrence at
> 
>
https://build.chromium.org/p/chromium.chromiumos/builders/Linux%20ChromiumOS%...
> 
> Original issue's description:
> > predictors: Basic browsertest checks predictor learning.
> >
> > First stage for implementing ResourcePrefetchPredictor browsertests is to
check
> > that the prefetch database is really filled after some navigations in
browser.
> > Browsertest class provides convenient way to declare expected resources in
the
> > test and then checks expectations internally that allows to write readable
> > tests.
> >
> > BUG=650253
> >
> > Committed: https://crrev.com/e636e3bcc172f220dfadfd2121d238c9908185bb
> > Cr-Commit-Position: refs/heads/master@{#429966}
> 
> TBR=pasko@chromium.org,lizeb@chromium.org,alexilin@chromium.org
> # Skipping CQ checks because original CL landed less than 1 days ago.
> NOPRESUBMIT=true
> NOTREECHECKS=true
> NOTRY=true
> BUG=650253
> 
> Committed: https://crrev.com/3250e71949fa7d226e9c843ee28277b988306229
> Cr-Commit-Position: refs/heads/master@{#430137}

BUG=650253
==========

[alexilin, 2016-11-07 17:04:28]

Description was changed from

==========
Reland of predictors: Basic browsertest checks predictor learning.

Reason for revert:
The reason of the test flakiness was fixed.

ResourcePrefetchPredictor needs to be initialized before learn any data.
Initialization starts when the first main frame load complete for given profile.
In our case it was about:blank page after tab creation. 
This initialization requires work in DB thread to be done. So there was a race
condition in which next navigation could appear earlier than initialization in
DB thread will be done.

For now we explicitly check that initialization of ResourcePrefetchPredictor was
done before any navigation.

Original issue's description:
> Revert of predictors: Basic browsertest checks predictor learning. (patchset
#9 id:160001 of https://codereview.chromium.org/2449323005/ )
> 
> Reason for revert:
> Created a flaky test: first occurrence at
> 
>
https://build.chromium.org/p/chromium.chromiumos/builders/Linux%20ChromiumOS%...
> 
> Original issue's description:
> > predictors: Basic browsertest checks predictor learning.
> >
> > First stage for implementing ResourcePrefetchPredictor browsertests is to
check
> > that the prefetch database is really filled after some navigations in
browser.
> > Browsertest class provides convenient way to declare expected resources in
the
> > test and then checks expectations internally that allows to write readable
> > tests.
> >
> > BUG=650253
> >
> > Committed: https://crrev.com/e636e3bcc172f220dfadfd2121d238c9908185bb
> > Cr-Commit-Position: refs/heads/master@{#429966}
> 
> TBR=pasko@chromium.org,lizeb@chromium.org,alexilin@chromium.org
> # Skipping CQ checks because original CL landed less than 1 days ago.
> NOPRESUBMIT=true
> NOTREECHECKS=true
> NOTRY=true
> BUG=650253
> 
> Committed: https://crrev.com/3250e71949fa7d226e9c843ee28277b988306229
> Cr-Commit-Position: refs/heads/master@{#430137}

BUG=650253
==========

to

==========
Reland of predictors: Basic browsertest checks predictor learning.

Reason for revert:
The reason of the test flakiness was fixed.

ResourcePrefetchPredictor needs to be initialized before learn any
data. Initialization starts when the first main frame load complete for
given profile. In our case it was about:blank page after tab creation.
This initialization requires work in DB thread to be done. So there was
a race condition in which next navigation could appear earlier than
initialization in DB thread will be done.

For now we explicitly check that initialization of
ResourcePrefetchPredictor was done before any navigation.

Original issue's description:
> Revert of predictors: Basic browsertest checks predictor learning. (patchset
#9 id:160001 of https://codereview.chromium.org/2449323005/ )
> 
> Reason for revert:
> Created a flaky test: first occurrence at
> 
>
https://build.chromium.org/p/chromium.chromiumos/builders/Linux%20ChromiumOS%...
> 
> Original issue's description:
> > predictors: Basic browsertest checks predictor learning.
> >
> > First stage for implementing ResourcePrefetchPredictor browsertests is to
check
> > that the prefetch database is really filled after some navigations in
browser.
> > Browsertest class provides convenient way to declare expected resources in
the
> > test and then checks expectations internally that allows to write readable
> > tests.
> >
> > BUG=650253
> >
> > Committed: https://crrev.com/e636e3bcc172f220dfadfd2121d238c9908185bb
> > Cr-Commit-Position: refs/heads/master@{#429966}
> 
> TBR=pasko@chromium.org,lizeb@chromium.org,alexilin@chromium.org
> # Skipping CQ checks because original CL landed less than 1 days ago.
> NOPRESUBMIT=true
> NOTREECHECKS=true
> NOTRY=true
> BUG=650253
> 
> Committed: https://crrev.com/3250e71949fa7d226e9c843ee28277b988306229
> Cr-Commit-Position: refs/heads/master@{#430137}

BUG=650253
==========

[alexilin, 2016-11-07 17:11:59]

Description was changed from

==========
Reland of predictors: Basic browsertest checks predictor learning.

Reason for revert:
The reason of the test flakiness was fixed.

ResourcePrefetchPredictor needs to be initialized before learn any
data. Initialization starts when the first main frame load complete for
given profile. In our case it was about:blank page after tab creation.
This initialization requires work in DB thread to be done. So there was
a race condition in which next navigation could appear earlier than
initialization in DB thread will be done.

For now we explicitly check that initialization of
ResourcePrefetchPredictor was done before any navigation.

Original issue's description:
> Revert of predictors: Basic browsertest checks predictor learning. (patchset
#9 id:160001 of https://codereview.chromium.org/2449323005/ )
> 
> Reason for revert:
> Created a flaky test: first occurrence at
> 
>
https://build.chromium.org/p/chromium.chromiumos/builders/Linux%20ChromiumOS%...
> 
> Original issue's description:
> > predictors: Basic browsertest checks predictor learning.
> >
> > First stage for implementing ResourcePrefetchPredictor browsertests is to
check
> > that the prefetch database is really filled after some navigations in
browser.
> > Browsertest class provides convenient way to declare expected resources in
the
> > test and then checks expectations internally that allows to write readable
> > tests.
> >
> > BUG=650253
> >
> > Committed: https://crrev.com/e636e3bcc172f220dfadfd2121d238c9908185bb
> > Cr-Commit-Position: refs/heads/master@{#429966}
> 
> TBR=pasko@chromium.org,lizeb@chromium.org,alexilin@chromium.org
> # Skipping CQ checks because original CL landed less than 1 days ago.
> NOPRESUBMIT=true
> NOTREECHECKS=true
> NOTRY=true
> BUG=650253
> 
> Committed: https://crrev.com/3250e71949fa7d226e9c843ee28277b988306229
> Cr-Commit-Position: refs/heads/master@{#430137}

BUG=650253
==========

to

==========
Reland of predictors: Basic browsertest checks predictor learning.

Reason for revert:
The reason of the test flakiness was fixed.

ResourcePrefetchPredictor needs to be initialized before learn any
data. Initialization starts when the first main frame load complete for
given profile. In our case it was about:blank page after tab creation.
This initialization requires work in DB thread to be done. So there was
a race condition in which next navigation could appear earlier than
initialization in DB thread will be done.
This bug could be easily reproduced by adding sleep in
ResourcePrefetchPredictorTables::GetAllData function.

For now we explicitly check that initialization of
ResourcePrefetchPredictor was done before any navigation.

Original issue's description:
> Revert of predictors: Basic browsertest checks predictor learning. (patchset
#9 id:160001 of https://codereview.chromium.org/2449323005/ )
> 
> Reason for revert:
> Created a flaky test: first occurrence at
> 
>
https://build.chromium.org/p/chromium.chromiumos/builders/Linux%20ChromiumOS%...
> 
> Original issue's description:
> > predictors: Basic browsertest checks predictor learning.
> >
> > First stage for implementing ResourcePrefetchPredictor browsertests is to
check
> > that the prefetch database is really filled after some navigations in
browser.
> > Browsertest class provides convenient way to declare expected resources in
the
> > test and then checks expectations internally that allows to write readable
> > tests.
> >
> > BUG=650253
> >
> > Committed: https://crrev.com/e636e3bcc172f220dfadfd2121d238c9908185bb
> > Cr-Commit-Position: refs/heads/master@{#429966}
> 
> TBR=pasko@chromium.org,lizeb@chromium.org,alexilin@chromium.org
> # Skipping CQ checks because original CL landed less than 1 days ago.
> NOPRESUBMIT=true
> NOTREECHECKS=true
> NOTRY=true
> BUG=650253
> 
> Committed: https://crrev.com/3250e71949fa7d226e9c843ee28277b988306229
> Cr-Commit-Position: refs/heads/master@{#430137}

BUG=650253
==========

[alexilin, 2016-11-07 17:13:56]

Description was changed from

==========
Reland of predictors: Basic browsertest checks predictor learning.

Reason for revert:
The reason of the test flakiness was fixed.

ResourcePrefetchPredictor needs to be initialized before learn any
data. Initialization starts when the first main frame load complete for
given profile. In our case it was about:blank page after tab creation.
This initialization requires work in DB thread to be done. So there was
a race condition in which next navigation could appear earlier than
initialization in DB thread will be done.
This bug could be easily reproduced by adding sleep in
ResourcePrefetchPredictorTables::GetAllData function.

For now we explicitly check that initialization of
ResourcePrefetchPredictor was done before any navigation.

Original issue's description:
> Revert of predictors: Basic browsertest checks predictor learning. (patchset
#9 id:160001 of https://codereview.chromium.org/2449323005/ )
> 
> Reason for revert:
> Created a flaky test: first occurrence at
> 
>
https://build.chromium.org/p/chromium.chromiumos/builders/Linux%20ChromiumOS%...
> 
> Original issue's description:
> > predictors: Basic browsertest checks predictor learning.
> >
> > First stage for implementing ResourcePrefetchPredictor browsertests is to
check
> > that the prefetch database is really filled after some navigations in
browser.
> > Browsertest class provides convenient way to declare expected resources in
the
> > test and then checks expectations internally that allows to write readable
> > tests.
> >
> > BUG=650253
> >
> > Committed: https://crrev.com/e636e3bcc172f220dfadfd2121d238c9908185bb
> > Cr-Commit-Position: refs/heads/master@{#429966}
> 
> TBR=pasko@chromium.org,lizeb@chromium.org,alexilin@chromium.org
> # Skipping CQ checks because original CL landed less than 1 days ago.
> NOPRESUBMIT=true
> NOTREECHECKS=true
> NOTRY=true
> BUG=650253
> 
> Committed: https://crrev.com/3250e71949fa7d226e9c843ee28277b988306229
> Cr-Commit-Position: refs/heads/master@{#430137}

BUG=650253
==========

to

==========
Reland of predictors: Basic browsertest checks predictor learning.

Reason for revert:
The reason of the test flakiness was fixed.

ResourcePrefetchPredictor needs to be initialized before learn any
data. Initialization starts when the first main frame load complete for
given profile. In our case it was about:blank page after tab creation.
This initialization requires work in DB thread to be done. So there was
a race condition in which next navigation could appear earlier than
initialization in DB thread will be done.
This bug could be easily reproduced by adding sleep in
ResourcePrefetchPredictorTables::GetAllData function.

For now we explicitly check that initialization of
ResourcePrefetchPredictor was done before any navigation.

Original issue's description:
> Revert of predictors: Basic browsertest checks predictor learning. (patchset
#9 id:160001 of https://codereview.chromium.org/2449323005/ )
> 
> Reason for revert:
> Created a flaky test: first occurrence at
> 
>
https://build.chromium.org/p/chromium.chromiumos/builders/Linux%20ChromiumOS%...
> 
> Original issue's description:
> > predictors: Basic browsertest checks predictor learning.
> >
> > First stage for implementing ResourcePrefetchPredictor browsertests is to
check
> > that the prefetch database is really filled after some navigations in
browser.
> > Browsertest class provides convenient way to declare expected resources in
the
> > test and then checks expectations internally that allows to write readable
> > tests.
> >
> > BUG=650253
> >
> > Committed: https://crrev.com/e636e3bcc172f220dfadfd2121d238c9908185bb
> > Cr-Commit-Position: refs/heads/master@{#429966}
> 
> TBR=pasko@chromium.org,lizeb@chromium.org,alexilin@chromium.org
> # Skipping CQ checks because original CL landed less than 1 days ago.
> NOPRESUBMIT=true
> NOTREECHECKS=true
> NOTRY=true
> BUG=650253
> 
> Committed: https://crrev.com/3250e71949fa7d226e9c843ee28277b988306229
> Cr-Commit-Position: refs/heads/master@{#430137}

BUG=650253
==========

[alexilin, 2016-11-07 17:14:05]

alexilin@chromium.org changed reviewers:
- mathp@chromium.org

[Benoit L, 2016-11-07 17:17:07]

https://codereview.chromium.org/2482833002/diff/170001/chrome/browser/predict...
File chrome/browser/predictors/resource_prefetch_predictor_browsertest.cc
(right):

https://codereview.chromium.org/2482833002/diff/170001/chrome/browser/predict...
chrome/browser/predictors/resource_prefetch_predictor_browsertest.cc:113:
EnsurePredictorInitialized();
This would be better in SetUp(), since we're going to need it everywhere, and
this is actually test setup, not part of the test.

[alexilin, 2016-11-07 17:17:12]

Hi all!

I've found problem in browser tests for ResourcePrefetchPredictor. The issue
description should be enough informative.

The only thing that we should be sure in the way of fix.

[alexilin, 2016-11-07 17:19:51]

https://codereview.chromium.org/2482833002/diff/170001/chrome/browser/predict...
File chrome/browser/predictors/resource_prefetch_predictor_browsertest.cc
(right):

https://codereview.chromium.org/2482833002/diff/170001/chrome/browser/predict...
chrome/browser/predictors/resource_prefetch_predictor_browsertest.cc:113:
EnsurePredictorInitialized();
On 2016/11/07 17:17:06, Benoit L wrote:
> This would be better in SetUp(), since we're going to need it everywhere, and
> this is actually test setup, not part of the test.

Done.

[Benoit L, 2016-11-07 17:23:35]

On 2016/11/07 17:19:51, alexilin wrote:
>
https://codereview.chromium.org/2482833002/diff/170001/chrome/browser/predict...
> File chrome/browser/predictors/resource_prefetch_predictor_browsertest.cc
> (right):
> 
>
https://codereview.chromium.org/2482833002/diff/170001/chrome/browser/predict...
> chrome/browser/predictors/resource_prefetch_predictor_browsertest.cc:113:
> EnsurePredictorInitialized();
> On 2016/11/07 17:17:06, Benoit L wrote:
> > This would be better in SetUp(), since we're going to need it everywhere,
and
> > this is actually test setup, not part of the test.
> 
> Done.

lgtm

[alexilin, 2016-11-07 17:43:15]

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

[commit-bot: I haz the power, 2016-11-07 17:43:57]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[pasko, 2016-11-07 19:09:36]

lgtm

nit: description: s/^Reason for revert:/Reason for reland:/

did you repro the flake by artificially slowing the initialization on the DB
thread?

[commit-bot: I haz the power, 2016-11-07 19:38:53]

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

[commit-bot: I haz the power, 2016-11-07 19:38:53]

Dry run: This issue passed the CQ dry run.

[alexilin, 2016-11-08 09:55:20]

Description was changed from

==========
Reland of predictors: Basic browsertest checks predictor learning.

Reason for revert:
The reason of the test flakiness was fixed.

ResourcePrefetchPredictor needs to be initialized before learn any
data. Initialization starts when the first main frame load complete for
given profile. In our case it was about:blank page after tab creation.
This initialization requires work in DB thread to be done. So there was
a race condition in which next navigation could appear earlier than
initialization in DB thread will be done.
This bug could be easily reproduced by adding sleep in
ResourcePrefetchPredictorTables::GetAllData function.

For now we explicitly check that initialization of
ResourcePrefetchPredictor was done before any navigation.

Original issue's description:
> Revert of predictors: Basic browsertest checks predictor learning. (patchset
#9 id:160001 of https://codereview.chromium.org/2449323005/ )
> 
> Reason for revert:
> Created a flaky test: first occurrence at
> 
>
https://build.chromium.org/p/chromium.chromiumos/builders/Linux%20ChromiumOS%...
> 
> Original issue's description:
> > predictors: Basic browsertest checks predictor learning.
> >
> > First stage for implementing ResourcePrefetchPredictor browsertests is to
check
> > that the prefetch database is really filled after some navigations in
browser.
> > Browsertest class provides convenient way to declare expected resources in
the
> > test and then checks expectations internally that allows to write readable
> > tests.
> >
> > BUG=650253
> >
> > Committed: https://crrev.com/e636e3bcc172f220dfadfd2121d238c9908185bb
> > Cr-Commit-Position: refs/heads/master@{#429966}
> 
> TBR=pasko@chromium.org,lizeb@chromium.org,alexilin@chromium.org
> # Skipping CQ checks because original CL landed less than 1 days ago.
> NOPRESUBMIT=true
> NOTREECHECKS=true
> NOTRY=true
> BUG=650253
> 
> Committed: https://crrev.com/3250e71949fa7d226e9c843ee28277b988306229
> Cr-Commit-Position: refs/heads/master@{#430137}

BUG=650253
==========

to

==========
Reland of predictors: Basic browsertest checks predictor learning.

Reason for reland:
The reason of the test flakiness was fixed.

ResourcePrefetchPredictor needs to be initialized before learn any
data. Initialization starts when the first main frame load complete for
given profile. In our case it was about:blank page after tab creation.
This initialization requires work in DB thread to be done. So there was
a race condition in which next navigation could appear earlier than
initialization in DB thread will be done.
This bug could be easily reproduced by adding sleep in
ResourcePrefetchPredictorTables::GetAllData function.

For now we explicitly check that initialization of
ResourcePrefetchPredictor was done before any navigation.

Original issue's description:
> Revert of predictors: Basic browsertest checks predictor learning. (patchset
#9 id:160001 of https://codereview.chromium.org/2449323005/ )
> 
> Reason for revert:
> Created a flaky test: first occurrence at
> 
>
https://build.chromium.org/p/chromium.chromiumos/builders/Linux%20ChromiumOS%...
> 
> Original issue's description:
> > predictors: Basic browsertest checks predictor learning.
> >
> > First stage for implementing ResourcePrefetchPredictor browsertests is to
check
> > that the prefetch database is really filled after some navigations in
browser.
> > Browsertest class provides convenient way to declare expected resources in
the
> > test and then checks expectations internally that allows to write readable
> > tests.
> >
> > BUG=650253
> >
> > Committed: https://crrev.com/e636e3bcc172f220dfadfd2121d238c9908185bb
> > Cr-Commit-Position: refs/heads/master@{#429966}
> 
> TBR=pasko@chromium.org,lizeb@chromium.org,alexilin@chromium.org
> # Skipping CQ checks because original CL landed less than 1 days ago.
> NOPRESUBMIT=true
> NOTREECHECKS=true
> NOTRY=true
> BUG=650253
> 
> Committed: https://crrev.com/3250e71949fa7d226e9c843ee28277b988306229
> Cr-Commit-Position: refs/heads/master@{#430137}

BUG=650253
==========

[alexilin, 2016-11-08 09:56:00]

The CQ bit was checked by alexilin@chromium.org

[alexilin, 2016-11-08 09:56:01]

The patchset sent to the CQ was uploaded after l-g-t-m from lizeb@chromium.org
Link to the patchset: https://codereview.chromium.org/2482833002/#ps200001
(title: "Rebase.")

[commit-bot: I haz the power, 2016-11-08 09:56:17]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-11-08 10:03:02]

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

[commit-bot: I haz the power, 2016-11-08 10:03:03]

Failed to apply the patch.
On branch working_branch
Your branch is up-to-date with 'origin/refs/pending/heads/master'.
nothing to commit, working tree clean

[commit-bot: I haz the power, 2016-11-08 10:03:25]

Description was changed from

==========
Reland of predictors: Basic browsertest checks predictor learning.

Reason for reland:
The reason of the test flakiness was fixed.

ResourcePrefetchPredictor needs to be initialized before learn any
data. Initialization starts when the first main frame load complete for
given profile. In our case it was about:blank page after tab creation.
This initialization requires work in DB thread to be done. So there was
a race condition in which next navigation could appear earlier than
initialization in DB thread will be done.
This bug could be easily reproduced by adding sleep in
ResourcePrefetchPredictorTables::GetAllData function.

For now we explicitly check that initialization of
ResourcePrefetchPredictor was done before any navigation.

Original issue's description:
> Revert of predictors: Basic browsertest checks predictor learning. (patchset
#9 id:160001 of https://codereview.chromium.org/2449323005/ )
> 
> Reason for revert:
> Created a flaky test: first occurrence at
> 
>
https://build.chromium.org/p/chromium.chromiumos/builders/Linux%20ChromiumOS%...
> 
> Original issue's description:
> > predictors: Basic browsertest checks predictor learning.
> >
> > First stage for implementing ResourcePrefetchPredictor browsertests is to
check
> > that the prefetch database is really filled after some navigations in
browser.
> > Browsertest class provides convenient way to declare expected resources in
the
> > test and then checks expectations internally that allows to write readable
> > tests.
> >
> > BUG=650253
> >
> > Committed: https://crrev.com/e636e3bcc172f220dfadfd2121d238c9908185bb
> > Cr-Commit-Position: refs/heads/master@{#429966}
> 
> TBR=pasko@chromium.org,lizeb@chromium.org,alexilin@chromium.org
> # Skipping CQ checks because original CL landed less than 1 days ago.
> NOPRESUBMIT=true
> NOTREECHECKS=true
> NOTRY=true
> BUG=650253
> 
> Committed: https://crrev.com/3250e71949fa7d226e9c843ee28277b988306229
> Cr-Commit-Position: refs/heads/master@{#430137}

BUG=650253
==========

to

==========
Reland of predictors: Basic browsertest checks predictor learning.

Reason for reland:
The reason of the test flakiness was fixed.

ResourcePrefetchPredictor needs to be initialized before learn any
data. Initialization starts when the first main frame load complete for
given profile. In our case it was about:blank page after tab creation.
This initialization requires work in DB thread to be done. So there was
a race condition in which next navigation could appear earlier than
initialization in DB thread will be done.
This bug could be easily reproduced by adding sleep in
ResourcePrefetchPredictorTables::GetAllData function.

For now we explicitly check that initialization of
ResourcePrefetchPredictor was done before any navigation.

Original issue's description:
> Revert of predictors: Basic browsertest checks predictor learning. (patchset
#9 id:160001 of https://codereview.chromium.org/2449323005/ )
>
> Reason for revert:
> Created a flaky test: first occurrence at
>
>
https://build.chromium.org/p/chromium.chromiumos/builders/Linux%20ChromiumOS%...
>
> Original issue's description:
> > predictors: Basic browsertest checks predictor learning.
> >
> > First stage for implementing ResourcePrefetchPredictor browsertests is to
check
> > that the prefetch database is really filled after some navigations in
browser.
> > Browsertest class provides convenient way to declare expected resources in
the
> > test and then checks expectations internally that allows to write readable
> > tests.
> >
> > BUG=650253
> >
> > Committed: https://crrev.com/e636e3bcc172f220dfadfd2121d238c9908185bb
> > Cr-Commit-Position: refs/heads/master@{#429966}
>
> TBR=pasko@chromium.org,lizeb@chromium.org,alexilin@chromium.org
> # Skipping CQ checks because original CL landed less than 1 days ago.
> NOPRESUBMIT=true
> NOTREECHECKS=true
> NOTRY=true
> BUG=650253
>
> Committed: https://crrev.com/3250e71949fa7d226e9c843ee28277b988306229
> Cr-Commit-Position: refs/heads/master@{#430137}

BUG=650253

Committed: https://crrev.com/f2d707775a8184e7bd18d2334ff2a679f3a90439
Cr-Commit-Position: refs/heads/master@{#430561}
==========

[commit-bot: I haz the power, 2016-11-08 10:03:26]

Patchset 4 (id:??) landed as
https://crrev.com/f2d707775a8184e7bd18d2334ff2a679f3a90439
Cr-Commit-Position: refs/heads/master@{#430561}

[alexilin, 2016-11-08 10:06:52]

pasko@: yes, I was able to reproduce this bug by slowing task in DB thread. I
suspected that this DB thread blocking could prevent some other browser
initialization needed to run navigation but no such thing.
Unittest for predictor also contains the similar initialization workaround thus
I had to be more careful with it in the browsertest.
https://cs.chromium.org/chromium/src/chrome/browser/predictors/resource_prefe...