ThreadRestrictions: disallow blocking IO on the UI thread

ThreadRestrictions: disallow blocking IO on the UI thread

This patch sets the "disallow IO" flag after the UI thread has
started, and then whitelists in the many places where we're
accidentally doing IO from the UI thread.  (I've filed bugs
on all of those cases.)

BUG=59847, 60630, 60641, 60211, 60634, 60643, 24163, 60825
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=63986

[jam, 2010-10-26 23:55:32]

lgtm

[willchan no longer on Chromium, 2010-10-27 00:06:18]

LGTM2

http://codereview.chromium.org/4146004/diff/3001/4001
File base/nss_util.cc (right):

http://codereview.chromium.org/4146004/diff/3001/4001#newcode316
base/nss_util.cc:316: //  
http://code.google.com/p/chromium/issues/detail?id=59847
The referenced bug is for this happening on the IO thread, which disagrees with
your changelist description.  if this happens on the IO thread, then that's
fine, it's a known bug and it'll be hard to work around.  If it's the UI thread,
then I'm a bit more disturbed.  Can you clarify?

A more anal approach would be to add the ScopedAllowIO higher up so we aren't
allowed to call EnsureNSSInit() on both the UI and IO thread, just whatever code
is calling this from the IO thread.

[Evan Martin, 2010-10-27 00:19:34]

On 2010/10/27 00:06:18, willchan wrote:
> http://codereview.chromium.org/4146004/diff/3001/4001#newcode316
> base/nss_util.cc:316: //  
> http://code.google.com/p/chromium/issues/detail?id=59847
> The referenced bug is for this happening on the IO thread, which disagrees
with
> your changelist description.  if this happens on the IO thread, then that's
> fine, it's a known bug and it'll be hard to work around.  If it's the UI
thread,
> then I'm a bit more disturbed.  Can you clarify?

Hm, I was hitting it on the UI thread.  I should've pasted in a call stack -- it
was something about setting up certificates IIRC.  I will track it down further.