Send resource load notifications from plugins to devtools agent

Send resource load notifications from plugins to devtools agent. This is a part of a fix which would allow to inspect plugin resources.

BUG=2084
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=33672

[yurys, 2009-11-26 15:16:40]

Plugins code needs to be changed to use 'unsigned long' for request id but
general idea should be clear.

[pfeldman, 2009-11-26 15:26:05]

Agent-related code looks good. Could you send this to the plugin folks so that
they could validate webframe_ access?

http://codereview.chromium.org/435044/diff/2001/2002
File webkit/glue/webplugin_impl.cc (right):

http://codereview.chromium.org/435044/diff/2001/2002#newcode881
webkit/glue/webplugin_impl.cc:881: if (!webframe_)
please be consistent about whether webframe_ can be null or not.

[yurys, 2009-11-26 15:43:02]

Guys, I picked your names from the changelog, please take a look at this patch
or forward it to a person who understands better what happens inside
WebPluginImpl.

[yurys, 2009-12-01 16:33:14]

ping

[Evan Martin, 2009-12-01 16:40:27]

I think John is the only person who can review this.  :)

I am a bit skeptical of the use of "unsigned long".  If you want a 64-bit
integer, it's safer to use "int64" directly; otherwise, you're getting a 32-bit
integer whether you use "long" or "int".

http://codereview.chromium.org/435044/diff/1003/5001
File webkit/glue/webplugin_impl.cc (right):

http://codereview.chromium.org/435044/diff/1003/5001#newcode787
webkit/glue/webplugin_impl.cc:787:
devtools_agent->didFailLoading(client_info->id, error);
2 space tab

[yurys, 2009-12-01 17:40:35]

On 2009/12/01 16:40:27, Evan Martin wrote:
> I think John is the only person who can review this.  :)
> 
> I am a bit skeptical of the use of "unsigned long".  If you want a 64-bit
> integer, it's safer to use "int64" directly; otherwise, you're getting a
32-bit
> integer whether you use "long" or "int".
> 
I don't think that unsigned long is the best choice but it is what
WebCore/loader code uses for request identifiers. sizeof(unsigned long) == 8 on
my x86_64 linux box so it depends on the compiler and not always 32-bit, though
there is no guarantee that it is 64-bit on 64-bit platform.

> http://codereview.chromium.org/435044/diff/1003/5001
> File webkit/glue/webplugin_impl.cc (right):
> 
> http://codereview.chromium.org/435044/diff/1003/5001#newcode787
> webkit/glue/webplugin_impl.cc:787:
> devtools_agent->didFailLoading(client_info->id, error);
> 2 space tab
Done.

[jam, 2009-12-01 18:25:07]

If this is changed to unsigned int, then all the other places that use it need
to be updated as well, i.e. these interfaces and the objects that implement them

WebPluginDelegate::CreateResourceClient(int resource_id
WebPlugin::SetDeferResourceLoading(int resource_id
WebPlugin::CancelResource(int id

[jam, 2009-12-01 18:26:30]

sorry for the delay, I missed this email.  In the future if I take this
long, feel free to email me directly or IM me.

On Thu, Nov 26, 2009 at 7:43 AM, <yurys@chromium.org> wrote:

> Guys, I picked your names from the changelog, please take a look at this
> patch
> or forward it to a person who understands better what happens inside
> WebPluginImpl.
>
>
> http://codereview.chromium.org/435044
>

[yurys, 2009-12-02 16:35:52]

Done. resource_id type has been changed to unsigned long everywhere in  plugins
code.

On 2009/12/01 18:25:07, John Abd-El-Malek wrote:
> If this is changed to unsigned int, then all the other places that use it need
> to be updated as well, i.e. these interfaces and the objects that implement
them
> 
> WebPluginDelegate::CreateResourceClient(int resource_id
> WebPlugin::SetDeferResourceLoading(int resource_id
> WebPlugin::CancelResource(int id

[jam, 2009-12-02 17:57:14]

lgtm

[dglazkov, 2009-12-03 16:56:23]

I am pretty sure this is causing a layout test to crash:

http://src.chromium.org/viewvc/chrome/trunk/src/webkit/tools/layout_tests/fla...