Avoid using NULL comparison to check if element is in the map.

Use 0, not NULL for member pointer comparision. Some gcc versions
print warning otherwise.

BUG=None
TEST=Compiles and works as expected

[Nikolay, 2012-09-14 12:17:38]

Some gcc versions (such as 4.4.6 used for ARM target) complains about member
pointer comparison with NULL.
http://stackoverflow.com/questions/4665709/is-it-always-valid-c-to-compare-a-...
has some discussion of this issue.

[dmichael (off chromium), 2012-09-17 15:10:35]

This is counter to the style guide, I think:
http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml#0_and_NULL
(Though admittedly it doesn't call out function or member pointers
specifically).

Is there a bug filed, or more detail on why you need to use a different GCC?
Since this is counter to the style guide, you're likely to get more warnings
appearing again later if there's not a bot building in this configuration.

[Nikolay, 2012-09-18 13:29:15]

I needed this compiler version, as it's the only crosstools gcc which
successfully builds Chrome on ARM. Not sure if it's against rules directly, as
NULL is usually declared as 0 is C++. However, if it is defined as (void*)0,
this comparison may be invalid, as we comparing member pointer to regular
pointer, and not sure if this is valid behavior. To avoid those problems, would
it be possible to rewrite code a bit, so that instead of null check of value
obtained form the map,
you check if there's value for given key first, and only then do lookup?

[dmichael (off chromium), 2012-09-18 15:09:17]

On 2012/09/18 13:29:15, olonho wrote:
> I needed this compiler version, as it's the only crosstools gcc which
> successfully builds Chrome on ARM. Not sure if it's against rules directly, as
> NULL is usually declared as 0 is C++. However, if it is defined as (void*)0,
> this comparison may be invalid, as we comparing member pointer to regular
> pointer, and not sure if this is valid behavior. To avoid those problems,
would
> it be possible to rewrite code a bit, so that instead of null check of value
> obtained form the map,
> you check if there's value for given key first, and only then do lookup?
Holy cow, yes, that's exactly what it should be doing. I didn't realize what was
happening there. It's std::map, so those "lookups" are inserting new items in to
the map. Should be relatively harmless, but it's still not right.

If you'd be willing to change those to proper lookups using find (or whatever
seems best), I'd appreciate it.

I guess we can sidestep having to violate the style guide this time. But I
agree, if it's a question of correctness over style, correctness has to win.

[Nikolay, 2012-09-19 10:27:52]

PTAL. BTW, which tree PPAPI changes shall be committed to?
I wrote this CL against ppapi in native_client tree, but unsure if this is
correct.

[dmichael (off chromium), 2012-09-19 14:57:07]

On 2012/09/19 10:27:52, olonho wrote:
> PTAL. BTW, which tree PPAPI changes shall be committed to?
> I wrote this CL against ppapi in native_client tree, but unsure if this is
> correct.
code looks good, but this code lives in the Chromium tree now. I thought the
patch looked a little funny. Once you have that patch ready, I'll be able to
stamp it quickly.

[Nikolay, 2012-09-20 09:20:33]

See https://codereview.chromium.org/10959008/ for change against Chromium tree.