String vector's comparison works wrong

String vector's comparison works wrong

The problem happened because SimpleClassVectorTraits was used as a vector trait for WTF::String class, hence 'memcmp' was used for string vector's comparison, which is apparently wrong.
This CL corrects vector trait for WTF::String and also introduces a unit test to verify that the problem is not happening any more.

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=157360

[Mikhail, 2013-09-06 07:38:56]

Could you please have a look?

[Chris Evans, 2013-09-06 08:02:00]

On 2013/09/06 07:38:56, mikhail.pozdnyakov wrote:
> Could you please have a look?

Interesting. String is has operator== so we expect it to be called, since its
behavior differs from memcmp(), is that a fair summary?

Note that the traits for Vector<RefPtr<T> > are currently also set to use
memcmp(), but that matches the behavior of RefPtr<T>::operator=(), is that also
your understanding?

[Chris Evans, 2013-09-06 08:10:08]

On 2013/09/06 08:02:00, Chris Evans wrote:
> On 2013/09/06 07:38:56, mikhail.pozdnyakov wrote:
> > Could you please have a look?
> 
> Interesting. String is has operator== so we expect it to be called, since its
> behavior differs from memcmp(), is that a fair summary?
> 
> Note that the traits for Vector<RefPtr<T> > are currently also set to use
> memcmp(), but that matches the behavior of RefPtr<T>::operator=(), is that
also
> your understanding?

LGTM, subject to tests passing of course :)

[Mikhail, 2013-09-06 08:13:01]

On 2013/09/06 08:02:00, Chris Evans wrote:
> On 2013/09/06 07:38:56, mikhail.pozdnyakov wrote:
> > Could you please have a look?
> 
> Interesting. String is has operator== so we expect it to be called, since its
> behavior differs from memcmp(), is that a fair summary?

yes operator==  is what we want to be called, memcmp() works fine with
AtomicStrings, but not with Strings as same strings can have pointers to
different StringImpl instances.
> 
> Note that the traits for Vector<RefPtr<T> > are currently also set to use
> memcmp(), but that matches the behavior of RefPtr<T>::operator=(), is that
also
> your understanding?

for comparing RefPtr<T> memcmp will definitely work, as equal RefPtrs should
contain same raw pointer.

[commit-bot: I haz the power, 2013-09-06 08:24:50]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/mikhail.pozdnyakov@intel.com/23819029/1

[commit-bot: I haz the power, 2013-09-06 09:42:14]

Change committed as 157360