clang/win: Let SK_TRACEHR not produce -Wunused-value warnings in release builds.

include/utils/win/SkHRESULT.h

Index: include/utils/win/SkHRESULT.h
diff --git a/include/utils/win/SkHRESULT.h b/include/utils/win/SkHRESULT.h
index 4f9d1bec796e11cd6612bd578f9fb2c143db8dd7..225d78079d00871034ec6077dc7e4e06f9a46828 100644
--- a/include/utils/win/SkHRESULT.h
+++ b/include/utils/win/SkHRESULT.h
@@ -16,7 +16,7 @@ void SkTraceHR(const char* file, unsigned long line,
 #ifdef SK_DEBUG
 #define SK_TRACEHR(_hr, _msg) SkTraceHR(__FILE__, __LINE__, _hr, _msg)
 #else
-#define SK_TRACEHR(_hr, _msg) _hr

[bungeman-skia, 2015/12/06 21:03:47]

I will say that this existing code is odd, in that SkTraceHR is void, but this
isn't (I suppose I wrote it, but that was a while ago). I think maybe the reason
why is to prevent other (real) unused warnings, in instances where SK_TRACEHR is
used independently of HR_GENERAL and the hresult will be reported (used) in
debug, but isn't actually used in release at all. This might happen when the
actual failure mode doesn't make a difference but one of the out parameters
already lets us know better if there was success. I think SkFontMgr_win_dw.cpp
does something like this.

Perhaps this should be 'sk_ignore_unused_variable(_hr)' instead.

[Nico, 2015/12/07 17:42:28]

Done. (It looks like skia isn't built with unused variable warnings enabled when
building it as part of Chromium; it built fine with PS1 for me. It still built
fine when I added "int a = 4;" to some random function in skia.)

+#define SK_TRACEHR(_hr, _msg)
 #endif
 
 #define HR_GENERAL(_ex, _msg, _ret) {\