Re: [ft] First pretest of FreeType 2.1.10

[Adam D. Moss]

Adam D. Moss wrote:
> [resend, as this hasn't turned up in over 24h]
> Werner LEMBERG wrote:
>
> > I've uploaded a first pretest version of 2.1.10!  This is mainly to
> > test whether the library compiles on various platforms, and to check
> > whether the library itself works correctly.  There are still some
> > rough edges with documentation files...
>
>
> Cool!  It works here (Linux/x86, static build, custom I/O module).
> Thanks for the hard work.
>
> Unfortunately I am seeing various regressions in stroking (which
> in 2.1.9 wasn't perfect, but was better).  I attach some pictures
> comparing 2.1.9 and 2.1.10.  This is FT_STROKER_LINECAP_BUTT /
> FT_STROKER_LINEJOIN_ROUND and VeraMono.ttf.  I can provide more
> detail if needed; similar but lesser regressions are visible in
> FT_STROKER_LINEJOIN_BEVEL.

Reverting just the following change in ftstroke.c
from 2.1.9->2.1.10 fixes all of my observed regressions:

@@ -822,9 +885,9 @@
     phi = stroker->angle_in + theta;

     thcos  = FT_Cos( theta );
-    sigma  = FT_MulFix( stroker->miter_limit, thcos );

-    if ( sigma < 0x10000L )
+    /* TODO: find better criterion to switch off the optimization */
+    if ( thcos < 0x4000 )
     {
       FT_Vector_From_Polar( &delta, stroker->radius,
                             stroker->angle_out + rotate );