[libfuzzer] moving icu fuzzers to //third_party/icu, additional fuzzers.

[libfuzzer] moving icu fuzzers to //third_party/icu, additional fuzzers.

Adding fuzzers for:
  - NumberFormat::parse
  - BreakIterator
  - ucasemap

BUG=539572

[aizatsky, 2016-05-18 22:05:40]

aizatsky@chromium.org changed reviewers:
+ jshin@chromium.org, mmoroz@chromium.org, ochang@chromium.org

[jungshik at Google, 2016-05-19 07:21:54]

Dependency on base::i18n is unfortunate, but for our internal use (as opposed to
upstreaming it), it should be fine. 

LGTM (my comments are about future improvements).

https://codereview.chromium.org/1995823002/diff/20001/BUILD.gn
File BUILD.gn (right):

https://codereview.chromium.org/1995823002/diff/20001/BUILD.gn#newcode54
BUILD.gn:54: }
I think these lines were added by another CL. Well, they'll go away when you
rebase :-)

https://codereview.chromium.org/1995823002/diff/20001/fuzzers/icu_break_itera...
File fuzzers/icu_break_iterator_fuzzer.cc (right):

https://codereview.chromium.org/1995823002/diff/20001/fuzzers/icu_break_itera...
fuzzers/icu_break_iterator_fuzzer.cc:14: icu::UnicodeString
str(reinterpret_cast<const char*>(data), size);
Again, this will test how well UnicodeString can deal with mostly invalid  input
strings but BreakIterator will not be  tested much.  (with valid input strings,
there can be hidden edge cases that may trigger a bug in BreakIterator). 

There should be a helper function to generate valid random strings in UTF-16 or
UTF-8. That way, One way would be to generate a sequence of random numbers ( [0,
0x10FFFF] ) and convert them to UTF-16 or UTF-8 strings.

https://codereview.chromium.org/1995823002/diff/20001/fuzzers/icu_ucasemap_fu...
File fuzzers/icu_ucasemap_fuzzer.cc (right):

https://codereview.chromium.org/1995823002/diff/20001/fuzzers/icu_ucasemap_fu...
fuzzers/icu_ucasemap_fuzzer.cc:37: case 0: ucasemap_utf8ToLower(csm.get(),
dst.get(), dst_size, src, size,
This will test how well ucasemap_utf8Foo can deal with invalid UTF-8 input (most
 of input will be invalid UTF-8), but will not test the code path for valid
input strings (much). 

You may as well consider testing ucase_* APIs accepting UTF-16 input
(reintepret_casting |data| to UChar*).

[aizatsky, 2016-05-19 20:57:42]

Submitted as
https://chromium.googlesource.com/chromium/deps/icu.git/+/06c914d938e16bb4779...

[mmoroz, 2016-05-20 10:35:37]

w00t, new fuzzers! Thanks Mike! LGTM