set user agent for content browser tests

set user agent for content browser tests

R=jam@chromium.org
BUG=90442
TEST=browser_tests AllUrlsApiTest.Whitelisted Extension passes w/ a component build of content.

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=107778

[Paweł Hajdan Jr., 2011-10-27 08:48:52]

Drive-by.

http://codereview.chromium.org/8402016/diff/1/content/test/browser_test_base.cc
File content/test/browser_test_base.cc (right):

http://codereview.chromium.org/8402016/diff/1/content/test/browser_test_base....
content/test/browser_test_base.cc:39: std::string product("Chrome/15.16.17.18");
Why not 0.0.0.0 as for DRT? Chrome/15 might suggest a real thing, I think it's
currently in stable.

http://codereview.chromium.org/8402016/diff/1/content/test/browser_test_base....
content/test/browser_test_base.cc:40:
webkit_glue::SetUserAgent(webkit_glue::BuildUserAgentFromProduct(product),
I don't like it anyway. We're calling BrowserMain, why not let it do it?
BrowserTestBase::SetUp should be kept extremely simple.

[Dirk Pranke, 2011-10-27 21:01:14]

On 2011/10/27 08:48:52, Paweł Hajdan Jr. wrote:
> Drive-by.
> 
>
http://codereview.chromium.org/8402016/diff/1/content/test/browser_test_base.cc
> File content/test/browser_test_base.cc (right):
> 
>
http://codereview.chromium.org/8402016/diff/1/content/test/browser_test_base....
> content/test/browser_test_base.cc:39: std::string
product("Chrome/15.16.17.18");
> Why not 0.0.0.0 as for DRT? Chrome/15 might suggest a real thing, I think it's
> currently in stable.
> 

No particular reason; content_shell uses 15.16.17.18 and I was copying that. I
can switch it to 0.0.0.0.

>
http://codereview.chromium.org/8402016/diff/1/content/test/browser_test_base....
> content/test/browser_test_base.cc:40:
> webkit_glue::SetUserAgent(webkit_glue::BuildUserAgentFromProduct(product),
> I don't like it anyway. We're calling BrowserMain, why not let it do it?
> BrowserTestBase::SetUp should be kept extremely simple.

BrowserMain is generic code and doesn't know what the user agent should be; it
needs to be specified by the embedder somehow. Normally that would retrieved
through the content client, but it appears that we don't use a content client
here, so I have to set the user agent explicitly.

[Dirk Pranke, 2011-10-27 21:02:06]

On 2011/10/27 21:01:14, Dirk Pranke wrote:
> On 2011/10/27 08:48:52, Paweł Hajdan Jr. wrote:
> > Drive-by.
> > 
> >
>
http://codereview.chromium.org/8402016/diff/1/content/test/browser_test_base.cc
> > File content/test/browser_test_base.cc (right):
> > 
> >
>
http://codereview.chromium.org/8402016/diff/1/content/test/browser_test_base....
> > content/test/browser_test_base.cc:39: std::string
> product("Chrome/15.16.17.18");
> > Why not 0.0.0.0 as for DRT? Chrome/15 might suggest a real thing, I think
it's
> > currently in stable.
> > 
> 
> No particular reason; content_shell uses 15.16.17.18 and I was copying that. I
> can switch it to 0.0.0.0.
> 
> >
>
http://codereview.chromium.org/8402016/diff/1/content/test/browser_test_base....
> > content/test/browser_test_base.cc:40:
> > webkit_glue::SetUserAgent(webkit_glue::BuildUserAgentFromProduct(product),
> > I don't like it anyway. We're calling BrowserMain, why not let it do it?
> > BrowserTestBase::SetUp should be kept extremely simple.
> 
> BrowserMain is generic code and doesn't know what the user agent should be; it
> needs to be specified by the embedder somehow. Normally that would retrieved
> through the content client, but it appears that we don't use a content client
> here, so I have to set the user agent explicitly.

I should've noted that I agree that it's lame to do this here, but I don't think
we currently have a better option. I am open to other suggestions.

[Dirk Pranke, 2011-10-28 00:42:46]

Okay, turns out my earlier diagnosis was dead wrong, and I think I must've been
confused by some combination of debugging w/ --single-process and not.

At any rate, browser_test_base.cc is fine; the content_client is set during the
test suite setup, and the problem is totally different ... when we were calling
the about: handler, we were picking up the instance of webkit_glue linked into
browser_tests, and not the instance picked up in content. This would be a
problem in chrome as well since we'd get the version of glue linked into
chrome.dll.

Fortunately, this is an easy fix.

[jam, 2011-10-28 01:12:28]

lgtm

[Paweł Hajdan Jr., 2011-10-28 08:40:18]

This looks much better to me, LGTM, thanks.