Send the index in the canonical list over IPC when using the OOP plugin loader.

Send the index in the canonical list over IPC when using the OOP plugin loader.

Comparing paths does not work all the time on Linux if the plugin is wrapped and
is initially loaded from a symlinked location.

BUG=17863
TEST=Waterfall Linux(dbg)(2) plugin_tests

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=108306

[Bernhard Bauer, 2011-11-02 16:03:51]

LGTM, with a suggestion though:

http://codereview.chromium.org/8440042/diff/1/content/utility/utility_thread_...
File content/utility/utility_thread_impl.cc (right):

http://codereview.chromium.org/8440042/diff/1/content/utility/utility_thread_...
content/utility/utility_thread_impl.cc:142: Send(new
UtilityHostMsg_LoadedPlugin(i, group->web_plugin_infos().front()));
I still feel a little bit silly just sending an increasing index back. I guess
we could send the original path instead of the index?

[jam, 2011-11-02 16:09:29]

lgtm

http://codereview.chromium.org/8440042/diff/1/content/common/utility_messages.h
File content/common/utility_messages.h (right):

http://codereview.chromium.org/8440042/diff/1/content/common/utility_messages...
content/common/utility_messages.h:73: size_t /* index in the vector */,
here and below, it's slightly better to send int instead of size_t over IPC. The
reason is if you ever want to talk between a 64 bit browser and 32 bit plugin
loader, the size_t's would be different sizes

[Bernhard Bauer, 2011-11-02 16:13:47]

http://codereview.chromium.org/8440042/diff/1/content/common/utility_messages.h
File content/common/utility_messages.h (right):

http://codereview.chromium.org/8440042/diff/1/content/common/utility_messages...
content/common/utility_messages.h:73: size_t /* index in the vector */,
On 2011/11/02 16:09:29, John Abd-El-Malek wrote:
> here and below, it's slightly better to send int instead of size_t over IPC.
The
> reason is if you ever want to talk between a 64 bit browser and 32 bit plugin
> loader, the size_t's would be different sizes

Oooh. Out of curiosity, what would be necessary to spawn a full 32-bit plugin
process for a 64-bit browser (or vice-versa)? Being able to run a 32-bit Flash
without nspluginwrapper would be nice.

[Robert Sesek, 2011-11-02 16:26:28]

Landing…

http://codereview.chromium.org/8440042/diff/1/content/common/utility_messages.h
File content/common/utility_messages.h (right):

http://codereview.chromium.org/8440042/diff/1/content/common/utility_messages...
content/common/utility_messages.h:73: size_t /* index in the vector */,
On 2011/11/02 16:09:29, John Abd-El-Malek wrote:
> here and below, it's slightly better to send int instead of size_t over IPC.
The
> reason is if you ever want to talk between a 64 bit browser and 32 bit plugin
> loader, the size_t's would be different sizes

Done. Used uint32_t because I'm comparing against size_t.

http://codereview.chromium.org/8440042/diff/1/content/utility/utility_thread_...
File content/utility/utility_thread_impl.cc (right):

http://codereview.chromium.org/8440042/diff/1/content/utility/utility_thread_...
content/utility/utility_thread_impl.cc:142: Send(new
UtilityHostMsg_LoadedPlugin(i, group->web_plugin_infos().front()));
On 2011/11/02 16:03:52, Bernhard Bauer wrote:
> I still feel a little bit silly just sending an increasing index back. I guess
> we could send the original path instead of the index?

I think the index is fine to send. Sending the original path is more data that
we likely wouldn't use.