Roll harfbuzz-ng to 0.9.40

Roll harfbuzz-ng to 0.9.40

Update the bundled version of harfbuzz-ng in third_party to 0.9.40 which
is the latest release. This includes a variety of fixes and also has the
two MacOS CoreText reverse_range fixes that we've already cherry picked.
TBR=behdad@chromium.org

Committed: https://crrev.com/745ff3a6e89a36a8f1b867e3875d352bf5f57505
Cr-Commit-Position: refs/heads/master@{#321873}

[eae, 2015-03-23 19:23:23]

eae@chromium.org changed reviewers:
+ behdad@chromium.org

[eae, 2015-03-23 19:23:29]

The CQ bit was checked by eae@chromium.org

[commit-bot: I haz the power, 2015-03-23 19:24:04]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1005693006/1

[commit-bot: I haz the power, 2015-03-23 22:23:18]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2015-03-23 22:23:51]

Patchset 1 (id:??) landed as
https://crrev.com/745ff3a6e89a36a8f1b867e3875d352bf5f57505
Cr-Commit-Position: refs/heads/master@{#321873}

[brucedawson, 2015-03-26 17:27:31]

Possible issues in hb-buffer-deserialize-text.hh?

I was investigating a Harfbuzz compiler warning (from VC++'s /analyze) and the
warning referred to a non-existent source file. That struck me as odd. I
eventually tracked this down to the #line directives that are in many Harfbuzz
header files. In particular, hb-buffer-deserialize-text.hh contains lines such
as:

    #line 43 "hb-buffer-deserialize-text.rl"

This causes odd warning paths such as:

    out\debug\hb-buffer-deserialize-text.rl(47) : warning C6001: Using
uninitialized memory 'pos'.

Note that VC++ has trusted the #line directive and then reported the path as if
the file was in the current directory. This is very confusing and I wanted to
suggest that the 144 #line directives in Harfbuzz should be removed, especially
since they refer to files that do not exist in the Chromium code base.

As for the warning, it occurs on line 504 of hb-buffer-deserialize-text.hh:

    buffer->pos[buffer->len - 1] = pos;

I don't know whether it is a legitimate warning. I think that in general it is
worthwhile to just initialize POD local variables when they are declared. That
makes it entirely a non-issue, and local variable initialization is extremely
cheap.

[behdad_google, 2015-03-26 18:15:11]

On 2015/03/26 17:27:31, brucedawson wrote:
> Possible issues in hb-buffer-deserialize-text.hh?
> 
> I was investigating a Harfbuzz compiler warning (from VC++'s /analyze) and the
> warning referred to a non-existent source file. That struck me as odd. I
> eventually tracked this down to the #line directives that are in many Harfbuzz
> header files. In particular, hb-buffer-deserialize-text.hh contains lines such
> as:
> 
>     #line 43 "hb-buffer-deserialize-text.rl"
> 
> This causes odd warning paths such as:
> 
>     out\debug\hb-buffer-deserialize-text.rl(47) : warning C6001: Using
> uninitialized memory 'pos'.
> 
> Note that VC++ has trusted the #line directive and then reported the path as
if
> the file was in the current directory. This is very confusing and I wanted to
> suggest that the 144 #line directives in Harfbuzz should be removed,
especially
> since they refer to files that do not exist in the Chromium code base.
> 
> As for the warning, it occurs on line 504 of hb-buffer-deserialize-text.hh:
> 
>     buffer->pos[buffer->len - 1] = pos;
> 
> I don't know whether it is a legitimate warning. I think that in general it is
> worthwhile to just initialize POD local variables when they are declared. That
> makes it entirely a non-issue, and local variable initialization is extremely
> cheap.

The warning is harmless.  But fixed anyway:
https://github.com/behdad/harfbuzz/commit/aee685086c8fde6f6c4590e483a177c19f2...

Thanks.