[PATCH] [libcxx] std::codecvt - short wchar support

[Marshall Clow]

I don't like the gutting of the tests.
I don't like the confusion added to the code.


REPOSITORY
  rL LLVM

================
Comment at: src/locale.cpp:1795
@@ -1868,2 +1794,3 @@
+utf16_to_utf8(const _FrmType* frm, const _FrmType* frm_end, const _FrmType*& frm_nxt,
               uint8_t* to, uint8_t* to_end, uint8_t*& to_nxt,
               unsigned long Maxcode = 0x10FFFF, codecvt_mode mode = codecvt_mode(0))
----------------
Same comment as below. 

================
Comment at: src/locale.cpp:1877
@@ -1946,2 +1876,3 @@
+              _ToType* to, _ToType* to_end, _ToType*& to_nxt,
               unsigned long Maxcode = 0x10FFFF, codecvt_mode mode = codecvt_mode(0))
 {
----------------
You're changing a function named `utf8_to_utf16` to take parameters of type `uint32_t *`?

Eww.  I object.

================
Comment at: test/std/localization/locales/locale.convenience/conversions/conversions.string/to_bytes.pass.cpp:46
@@ -35,2 +45,3 @@
         assert(bs == "\xF1\x80\x80\x83");
+#endif
         bs = myconv.to_bytes(L"");
----------------
This is not "fixing the libcxx tests dependent on std::codecvt".

This is *removing the tests*.

http://reviews.llvm.org/D8880

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/