Change BufferManager::RequestBufferAccess signatures to avoid sprintfs.

Change BufferManager::RequestBufferAccess signatures to avoid sprintfs.

Using varargs and va_list eliminates the cost of string construction
in the callers.

BUG=661186, 654201
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel

Committed: https://crrev.com/49f47242e1c13505573e4c469a5f61f7f65f886b
Cr-Commit-Position: refs/heads/master@{#429375}

[Ken Russell (switch to Gerrit), 2016-11-02 02:55:05]

Description was changed from

==========
Change BufferManager::RequestBufferAccess signatures to avoid sprintfs.

Using varargs and va_list eliminates the cost of string construction
in the callers.

BUG=661186, 654201
==========

to

==========
Change BufferManager::RequestBufferAccess signatures to avoid sprintfs.

Using varargs and va_list eliminates the cost of string construction
in the callers.

BUG=661186, 654201
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
==========

[Ken Russell (switch to Gerrit), 2016-11-02 02:58:06]

kbr@chromium.org changed reviewers:
+ piman@chromium.org, zmo@chromium.org

[Ken Russell (switch to Gerrit), 2016-11-02 02:58:07]

zmo or piman: please review. Thanks.

Please see https://bugs.chromium.org/p/chromium/issues/detail?id=661186#c9 for
profiles before and after this patch.

marcheu: FYI.

[piman, 2016-11-02 17:27:41]

https://codereview.chromium.org/2468123002/diff/1/gpu/command_buffer/service/...
File gpu/command_buffer/service/buffer_manager.cc (right):

https://codereview.chromium.org/2468123002/diff/1/gpu/command_buffer/service/...
gpu/command_buffer/service/buffer_manager.cc:818: std::string message_tag =
base::StringPrintV(error_message_format, varargs);
According to http://en.cppreference.com/w/cpp/utility/variadic/va_list, we need
to call va_end (and, I assume, va_start again?) if varargs was used in
RequestBufferAccessV called from l.811, before it's used again here in the
calling function.
TBH I'm not sure of the exact semantics of passing a va_list by value... This
scares me a bit.

At a high level - do we need that level of generality? This function is always
called with the same error_message_format, and the same next argument (target).
Could we instead pass the target through and use it to generate the message
("bound to target 0x%x") at the error location?

[Zhenyao Mo, 2016-11-02 17:31:18]

lgtm
Thanks for fixing this

[Zhenyao Mo, 2016-11-02 17:31:24]

The CQ bit was checked by zmo@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-11-02 17:32:12]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[Ken Russell (switch to Gerrit), 2016-11-02 18:29:01]

https://codereview.chromium.org/2468123002/diff/1/gpu/command_buffer/service/...
File gpu/command_buffer/service/buffer_manager.cc (right):

https://codereview.chromium.org/2468123002/diff/1/gpu/command_buffer/service/...
gpu/command_buffer/service/buffer_manager.cc:818: std::string message_tag =
base::StringPrintV(error_message_format, varargs);
On 2016/11/02 17:27:41, piman wrote:
> According to http://en.cppreference.com/w/cpp/utility/variadic/va_list, we
need
> to call va_end (and, I assume, va_start again?) if varargs was used in
> RequestBufferAccessV called from l.811, before it's used again here in the
> calling function.

If it was used in RequestBufferAccessV, that meant that line 811 returned false,
so va_end was called above and this function returned early.

> TBH I'm not sure of the exact semantics of passing a va_list by value... This
> scares me a bit.

It's legal to pass a va_list by value. I checked the Chromium sources first and
this is already done in several places where string formatting with varargs is
needed.

> At a high level - do we need that level of generality? This function is always
> called with the same error_message_format, and the same next argument
(target).
> Could we instead pass the target through and use it to generate the message
> ("bound to target 0x%x") at the error location?

This variant is called from the outside, but only in one place:
GLES2DecoderImpl::ValidateCompressedTexFuncData. It can be simplified. Thanks
for the suggestion. I've uploaded a new patch doing so.

https://codereview.chromium.org/2468123002/diff/20001/gpu/command_buffer/serv...
File gpu/command_buffer/service/buffer_manager.cc (right):

https://codereview.chromium.org/2468123002/diff/20001/gpu/command_buffer/serv...
gpu/command_buffer/service/buffer_manager.cc:872: bool
BufferManager::RequestBufferAccessV(ErrorState* error_state,
While there's only one caller of this now, it simplifies the code in my opinion
so I'd prefer to leave it in.

[piman, 2016-11-02 18:41:08]

lgtm

[Ken Russell (switch to Gerrit), 2016-11-02 18:57:07]

The CQ bit was checked by kbr@chromium.org

[Ken Russell (switch to Gerrit), 2016-11-02 18:57:07]

The patchset sent to the CQ was uploaded after l-g-t-m from zmo@chromium.org
Link to the patchset: https://codereview.chromium.org/2468123002/#ps20001
(title: "Simplified one RequestBufferAccess variant causing nested va_list
usage.")

[commit-bot: I haz the power, 2016-11-02 18:58:03]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-11-02 19:57:58]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2016-11-02 20:01:34]

Description was changed from

==========
Change BufferManager::RequestBufferAccess signatures to avoid sprintfs.

Using varargs and va_list eliminates the cost of string construction
in the callers.

BUG=661186, 654201
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
==========

to

==========
Change BufferManager::RequestBufferAccess signatures to avoid sprintfs.

Using varargs and va_list eliminates the cost of string construction
in the callers.

BUG=661186, 654201
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel

Committed: https://crrev.com/49f47242e1c13505573e4c469a5f61f7f65f886b
Cr-Commit-Position: refs/heads/master@{#429375}
==========

[commit-bot: I haz the power, 2016-11-02 20:01:35]

Patchset 2 (id:??) landed as
https://crrev.com/49f47242e1c13505573e4c469a5f61f7f65f886b
Cr-Commit-Position: refs/heads/master@{#429375}