commit f0285cde6e11a1178dbdb30a3503ae6603526c88
author Antonio Gomes <tonikitoo@igalia.com> Mon Jun 18 21:51:38 2018
committer Commit Bot <commit-bot@chromium.org> Mon Jun 18 21:51:38 2018
tree f9a08b749fbd6bf3dd15b727ecf9d7e4ae871cf3
parent f988ce79e2d72cd00817583def5a7b46177bc79a

Reland "Migrate ContentHashFetcher to SimpleURLLoader"

This is a reland of 27f8891c7cd9a86b2c92508cf6818726df4058d6

The main difference from the original CL is that instead of passing raw
network::mojom::URLLoaderFactory pointers, this CL takes a more conservative
approach where a thread-agnostic network::mojom::URLLoaderFactoryPtrInfo instance
is passed through ContentHash::FetchParams, which allows us to "bind" the
network::URLLoaderFactoryPtr instance only when it is actually going to be used
(in ContentHashFetcher::Start).

Also, differently than the original/reverted CL, this CL does not change
the threads ContentHash methods run on.

Last, in order to adapt ContentVerifierHashTest, we:

1) Replace the use of net::TestURLRequestInterceptor by content::URLLoaderInterceptor.
2) Install the interceptor at the beginning of each test execution flow, so
that it is able to intercept all relevant URLLoaderFactory creation requests.

Some other minor differences, were because of some preparation CLs already landed,
eg https://crrev.com/c/1089273.

Original change's description:
> Migrate ContentHashFetcher to SimpleURLLoader
>
> This CL migrates ContentHashFetcher away from URLRequestFetcher to
> SimpleURLLoader.
>
> In order to accomplish this, a few other related changes are also performed:
>
> 1) The regular flow of the code is ContentVerifier -> ContentHash ->
>    ContentHashFetcher. In ContentVerifier, for instance, UI, IO and
>    "File tasks" threads take place.
>    Previously, the URLRequestFetcher logic residing in ContentHashFetcher
>    executed in the "file tasks" thread.
>
>    As part of the migration from URLRequestFetcher to SimpleURLLoader
>    machinery, this CL also changes the ContentHashFetcher logic to
>    execute to the IO thread.
>    Note that it could be possible to keep the new SimpleURLLoader logic
>    in the "file tasks" thread. However, this would impose a way bigger change,
>    and require unittests to be considerably rewritten as well (see for
>    patchsets 7, 8 and 9.
>    The issue is that mojo objects are not thread safe. We could create a new
>    URLLoaderFactory for the extensions thread, but then we'd need to create a
>    new one when the network service crashes (Something this SharedURLLoaderFactory
>    one, already bound to the IOThread, magically handles), but that would be
>    significantly more complicated.
>
> 2) content_verifier_hash_fetch_behavior_browsertest.cc was changed
>    to support running with the network service feature both enabled
>    and disabled. This is a pattern that is also present in various other
>    unittests.
>
> In summary, the migration to use SimpleURLLoader performed here includes
> also a change in the thread that ContentHashFetcher runs on, but no
> functionality change is expected from this CL.
>
> BUG=773295,844926
>
> Change-Id: If570f69d01ff75ac59d8d043f8687621336dddcf
> Reviewed-on: https://chromium-review.googlesource.com/1056587
> Commit-Queue: Antonio Gomes <tonikitoo@igalia.com>
> Reviewed-by: Devlin <rdevlin.cronin@chromium.org>
> Reviewed-by: Matt Menke <mmenke@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#561480}

Bug: 773295, 844926
Cq-Include-Trybots: luci.chromium.try:linux_mojo
Change-Id: Ie2f3665bcca5378ff67bf78bec38bdc225ff6c26
Reviewed-on: https://chromium-review.googlesource.com/1076947
Commit-Queue: Antonio Gomes <tonikitoo@igalia.com>
Reviewed-by: Devlin <rdevlin.cronin@chromium.org>
Reviewed-by: Matt Menke <mmenke@chromium.org>
Cr-Commit-Position: refs/heads/master@{#568178}