Add a comment explaining why we need libxml's multithreading mode

Add a comment explaining why we need libxml's multithreading mode

BUG=138571
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=151111

[Peter Beverloo, 2012-08-08 20:59:41]

SGTM for Android.

What is the reason for disabling threading? Attaching a bug or a clearer
description of the why would be helpful.

[Nico, 2012-08-09 00:04:14]

Mark added -D_REENTRANT in the initial libxml.gypi, Peter added it for android.
When Peter added it, Evan wondered in the review if we need it at all. The bots
seem happy without it. (And Peter is happy with removing it too.)

So, can we just do this change? The motivation is fixing a compiler warning in
the threading code in libxml, which is admittedly a silly reason. However, this
seems like a change we'd want to make just for its own sake too.

[Mark Mentovai, 2012-08-09 15:42:42]

LGTM. I only had it threaded because that’s what a normal “configure” gave.

[Evan Martin, 2012-08-09 18:27:37]

On 2012/08/09 15:42:42, Mark Mentovai wrote:
> LGTM. I only had it threaded because that’s what a normal “configure” gave.

It seems you need threads if you use libxml at all from multiple threads:
http://www.xmlsoft.org/threads.html
That is, this flag is *not* related to the more specific and unChromelike thing
of touching the same xml document from multiple threads.

It seems to me likely (and/or possible) that Chrome touches libxml from more
than one thread, so maybe this change is a bad idea.

[Nico, 2012-08-10 18:13:00]

On Thu, Aug 9, 2012 at 11:27 AM,  <evan@chromium.org> wrote:
> On 2012/08/09 15:42:42, Mark Mentovai wrote:
>>
>> LGTM. I only had it threaded because that’s what a normal “configure”
>> gave.
>
>
> It seems you need threads if you use libxml at all from multiple threads:
> http://www.xmlsoft.org/threads.html
> That is, this flag is *not* related to the more specific and unChromelike
> thing
> of touching the same xml document from multiple threads.
>
> It seems to me likely (and/or possible) that Chrome touches libxml from more
> than one thread, so maybe this change is a bad idea.

Thanks for pointing this out. I repurposed this CL to add a comment to
that end instead.

(Unfortunate!)