Defalke testOnPageFinishedCalledAfterError

Defalke testOnPageFinishedCalledAfterError

by making timeout from 5 (default) to 15 seconds.

BUG=152033
Android java test change. Ran through android trybot
NOTRY=true

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=170848

[boliu, 2012-12-03 18:22:45]

See bug

[mkosiba (inactive), 2012-12-03 18:29:25]

LGTM

Thanks for de-flaking these!

[mkosiba (inactive), 2012-12-03 18:31:15]

btw. I have a feeling that this is somehow related to the network. If only 'bad
URL' tests are failing then maybe the amount of time required for this to fail
is determined by how long it takes for the dns query to fail?

[commit-bot: I haz the power, 2012-12-04 00:12:03]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/boliu@chromium.org/11316304/5002

[commit-bot: I haz the power, 2012-12-04 00:13:50]

Change committed as 170848

[boliu, 2012-12-04 08:20:19]

testOnPageFinishedCalledAfterError will pass with a connection error, so
suing localhost:7 is fine.
testOnReceivedErrorOnInvalidUrl explicitly checks for dns error, so won't
work with localhost:7. Any idea how to (more reliably) fake a dns error?


On Mon, Dec 3, 2012 at 2:50 PM, Jonathan Dixon <joth@chromium.org> wrote:

> Good point.
>
http://man.id.be.really.surprised.if.this.address.existed.blah<http://www.goo...
> /";
>
> Yeah, that's going to have painful failure modes due to the DNS timeout.
> Would "http://localhost:7/non_existant" work just as well?
>
>
>
>
>
>
>
> On 3 December 2012 10:31, <mkosiba@chromium.org> wrote:
>
>> btw. I have a feeling that this is somehow related to the network. If
>> only 'bad
>> URL' tests are failing then maybe the amount of time required for this to
>> fail
>> is determined by how long it takes for the dns query to fail?
>>
>>
https://codereview.chromium.**org/11316304/<https://codereview.chromium.org/1...
>>
>
>  --
>
>
>

[joth, 2012-12-04 08:42:35]

I'd be tempted to change it to a bad URL and then look for ERROR_BAD_URL
instead. (I don't think there's reason we absolutely must check DNS failure
error code, but no others?)


On 3 December 2012 15:05, Bo Liu <boliu@chromium.org> wrote:

> testOnReceivedErrorOnInvalidUrl

[mnaganov (inactive), 2012-12-04 10:05:37]

On Mon, Dec 3, 2012 at 11:05 PM, Bo Liu <boliu@chromium.org> wrote:
> testOnPageFinishedCalledAfterError will pass with a connection error, so
> suing localhost:7 is fine.
> testOnReceivedErrorOnInvalidUrl explicitly checks for dns error, so won't
> work with localhost:7. Any idea how to (more reliably) fake a dns error?
>

There is a fake_dns server in tools/android. It will be more reliable
to use it for serving DNS queries, as it runs locally. But I guess we
will need to teach it to emit error replies.

>
> On Mon, Dec 3, 2012 at 2:50 PM, Jonathan Dixon <joth@chromium.org> wrote:
>>
>> Good point.
>> http://man.id.be.really.surprised.if.this.address.existed.blah/";
>>
>> Yeah, that's going to have painful failure modes due to the DNS timeout.
>> Would "http://localhost:7/non_existant" work just as well?
>>
>>
>>
>>
>>
>>
>>
>> On 3 December 2012 10:31, <mkosiba@chromium.org> wrote:
>>>
>>> btw. I have a feeling that this is somehow related to the network. If
>>> only 'bad
>>> URL' tests are failing then maybe the amount of time required for this to
>>> fail
>>> is determined by how long it takes for the dns query to fail?
>>>
>>> https://codereview.chromium.org/11316304/
>>
>>
>> --
>>
>>
>
>
> --
>
>

[mkosiba (inactive), 2012-12-04 11:25:55]

the only reason this test checks for that specific error is because we have
a CTS test that does the same thing.
I wonder if we could somehow inject a mock dns resolver...


On Mon, Dec 3, 2012 at 11:55 PM, Jonathan Dixon <joth@chromium.org> wrote:

> I'd be tempted to change it to a bad URL and then look for ERROR_BAD_URL
> instead. (I don't think there's reason we absolutely must check DNS failure
> error code, but no others?)
>
>
> On 3 December 2012 15:05, Bo Liu <boliu@chromium.org> wrote:
>
>> testOnReceivedErrorOnInvalidUrl
>
>
>  --
>
>
>

[mnaganov (inactive), 2012-12-04 11:38:22]

You can write the resolver in Java. Another thing you need is to set
the system property "net.dns1" to use localhost (and restore it back
when the test ends).

On Tue, Dec 4, 2012 at 11:25 AM, Martin Kosiba <mkosiba@chromium.org> wrote:
> the only reason this test checks for that specific error is because we have
> a CTS test that does the same thing.
> I wonder if we could somehow inject a mock dns resolver...
>
>
> On Mon, Dec 3, 2012 at 11:55 PM, Jonathan Dixon <joth@chromium.org> wrote:
>>
>> I'd be tempted to change it to a bad URL and then look for ERROR_BAD_URL
>> instead. (I don't think there's reason we absolutely must check DNS failure
>> error code, but no others?)
>>
>>
>> On 3 December 2012 15:05, Bo Liu <boliu@chromium.org> wrote:
>>>
>>> testOnReceivedErrorOnInvalidUrl
>>
>>
>> --
>>
>>
>
>
> --
>
>

[mkosiba (inactive), 2012-12-04 13:59:22]

Oh, didn't know that. Thanks for the suggestion Mikhail!


On Tue, Dec 4, 2012 at 11:37 AM, Mikhail Naganov <mnaganov@chromium.org>wrote:

> You can write the resolver in Java. Another thing you need is to set
> the system property "net.dns1" to use localhost (and restore it back
> when the test ends).
>
> On Tue, Dec 4, 2012 at 11:25 AM, Martin Kosiba <mkosiba@chromium.org>
> wrote:
> > the only reason this test checks for that specific error is because we
> have
> > a CTS test that does the same thing.
> > I wonder if we could somehow inject a mock dns resolver...
> >
> >
> > On Mon, Dec 3, 2012 at 11:55 PM, Jonathan Dixon <joth@chromium.org>
> wrote:
> >>
> >> I'd be tempted to change it to a bad URL and then look for ERROR_BAD_URL
> >> instead. (I don't think there's reason we absolutely must check DNS
> failure
> >> error code, but no others?)
> >>
> >>
> >> On 3 December 2012 15:05, Bo Liu <boliu@chromium.org> wrote:
> >>>
> >>> testOnReceivedErrorOnInvalidUrl
> >>
> >>
> >> --
> >>
> >>
> >
> >
> > --
> >
> >
>

[joth, 2012-12-04 23:07:33]

Sounds to me then like we'll get better overall coverage by using a
different error code? :)
but mikhail's the dns tricks sound good. do cts use either those?

oh, for good measure chromium also has a way to add pre-canned DNS resolver
rules for test too. see net/base/mock_host_resolver.h



On 4 December 2012 03:25, Martin Kosiba <mkosiba@chromium.org> wrote:

> the only reason this test checks for that specific error is because we
> have a CTS test that does the same thing.
> I wonder if we could somehow inject a mock dns resolver...
>
>
> On Mon, Dec 3, 2012 at 11:55 PM, Jonathan Dixon <joth@chromium.org> wrote:
>
>> I'd be tempted to change it to a bad URL and then look for ERROR_BAD_URL
>> instead. (I don't think there's reason we absolutely must check DNS failure
>> error code, but no others?)
>>
>>
>> On 3 December 2012 15:05, Bo Liu <boliu@chromium.org> wrote:
>>
>>> testOnReceivedErrorOnInvalidUrl
>>
>>
>>  --
>>
>>
>>
>
>