add portable and canonical font support for DM

dm/DM.cpp

Index: dm/DM.cpp
diff --git a/dm/DM.cpp b/dm/DM.cpp
index 003145b92187a92743051a57bd1e5724612d5b2e..1c851334a17dfaa026357d5280ebe972370b3bf3 100644
--- a/dm/DM.cpp
+++ b/dm/DM.cpp
@@ -9,6 +9,8 @@
 #include "SkString.h"
 #include "Test.h"
 #include "gm.h"
+#include "sk_tool_utils.h"
+#include "sk_tool_utils_flags.h"
 
 #include "DMCpuGMTask.h"
 #include "DMGpuGMTask.h"
@@ -48,6 +50,8 @@ DEFINE_string(skps, "", "Directory to read skps from.");
 
 DEFINE_bool(gms, true, "Run GMs?");
 DEFINE_bool(tests, true, "Run tests?");
+DEFINE_bool(reportUsedChars, false, "Output test font construction data to be pasted into"

[mtklein, 2014/07/30 15:03:57]

Is it silly to just preemptively create paths for all characters we're likely to
type?

[caryclark, 2014/07/30 15:45:32]

My first pass was to do just that, and that may make the most sense once the GMs
are tightened up. 

My constraint for this CL is to leave the GMs alone -- I didn't want to confuse
this CL with rebaselining that might be caused by mistakes versus intentional
changes to make GMs better. So that required that I support all of the fonts and
chars that the current GMs use.

Enlarging these font ranges to create paths for all ASCII ballooned up the file
that creates the paths to over 1M, which means it can no longer be uploaded and
try-botted. For now, I'd rather have a just-large enough file that at least can
be part of this CL.

+                                    " create_test_font.cpp.");
 
 __SK_FORCE_IMAGE_DECODER_LINKING;
 
@@ -231,7 +235,13 @@ int dm_main() {
     tasks.wait();
 
     SkDebugf("\n");
-
+#ifdef SK_DEBUG

[mtklein, 2014/07/30 15:03:57]

Seems like you can drop the #ifdef/#endif wrapping?

[caryclark, 2014/07/30 15:45:32]

The report_used_chars relies on data gathered only in SK_DEBUG. So, while I
could drop the wrapping here, it would mislead the reader possibly to believe
that it is possible to report used chars in Release.

+    if (FLAGS_portableFonts && FLAGS_reportUsedChars) {
+        sk_tool_utils::report_used_chars();
+    }
+#endif
+    sk_tool_utils::release_portable_typeface();

[mtklein, 2014/07/30 15:03:57]

Is there a corresponding ref/acquire?  How do we know when we need to call this?
 Can it become part of SkGraphics?

[caryclark, 2014/07/30 15:45:32]

I don't doubt this could be better. Would this be more clear?

if (FLAGS_portableFonts) {
    sk_tools_utils::release_portable_typeface();
}

The ref/acquire happens in sk_tool_utils_font.cpp, in create_font(). I'd prefer
not to put this in SkGraphics but to leave this as a tool that is outside the
src/core world.

[mtklein, 2014/07/30 17:19:05]

Think we could have the portable font code register release_portable_typeface()
to be called with atexit the first time we create one?  Just seems tricky to
have to know here that somewhere inside code called by DM we we may have created
one.  Am I thinking right that GM could be using these too if you turn on the
flags, but there's no way for it to clean them up?

Maybe tack on an "s" to "typeface" too?

[caryclark, 2014/07/30 19:00:16]

Done.

[caryclark, 2014/07/30 19:00:16]

Done.

+    
     SkTArray<SkString> failures;
     reporter.getFailures(&failures);
     report_failures(failures);