Trace memory records source function for posted tasks instead of "RunTask"

Trace memory records source function for posted tasks instead of "RunTask"

The chrome://tracing system for memory intentionally only records two pieces
of data for each TRACE_EVENT - the category and name of the event. This keeps
its data structures small (since tcmalloc reads them frequently during heap
profiling) and is all that is needed for binning most allocations into groups.

However, grouping all memory allocated by posted tasks under "RunTask" isn't
useful - the user needs to know which task actually resulted in the memory
allocations. This patch attributes the allocations to the source function of
the task, allowing the user to find the actual source.

BUG=trace-viewer:331
TEST=manual, run chrome://tracing with memory checked, click on a sample in the heap row, the breakdown shows individual function names instead of "RunTask"

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=219411

[James Cook, 2013-08-22 22:59:15]

dsinclair, would you feel comfortable reviewing this patch?  nduca is out, but I
could wait until he gets back if you prefer.

I could also implement this like I did in
https://codereview.chromium.org/15418002/diff/45001/base/message_loop.cc if you
would prefer not to add another macro in trace_event.h

[dsinclair, 2013-08-23 13:39:47]

https://codereview.chromium.org/22909045/diff/1/base/debug/trace_event.h
File base/debug/trace_event.h (right):

https://codereview.chromium.org/22909045/diff/1/base/debug/trace_event.h#newc...
base/debug/trace_event.h:242: category_group, name, arg1_name, arg1_val,
arg2_name, arg2_val)
I like the use of a macro over doing it was done in the other CL, I really don't
like the macro name or it's magic grabbing of a given argument. What about
something like:

TRACE_EVENT_MEMORY_REALLOCATED2(
  "task",
  "MessageLoop::RunTask", 
  pending_task.posted_from.function_name(),
  "src_file", pending_task.posted_from.file_name(),
  "src_func", pending_task.posted_from.function_name());


Something like that? Is that worse, heh. I like having the explicit third
argument that is the thing you want to rename memory too, I think that's a lot
clearer then realizing arg2 value is going to be used.

[James Cook, 2013-08-23 15:56:24]

dsinclair, PTAL

[dsinclair, 2013-08-23 15:59:35]

lgtm

[James Cook, 2013-08-23 16:05:50]

jar, can I get OWNERS for base/message_loop?

This is a small refinement to my chrome:tracing tcmalloc memory tracking.

[jar (doing other things), 2013-08-23 22:55:23]

base LGTM

[commit-bot: I haz the power, 2013-08-23 23:03:15]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/jamescook@chromium.org/22909045/6001

[commit-bot: I haz the power, 2013-08-24 03:26:45]

Change committed as 219411