Re: [Fontforge-devel] Several problems

[Don K. <aar...@kr...>]

On Tue, 2005-05-17 at 13:31, Don Koch wrote:
>> Are you not using the sourceforge bug tracking system?  If not, there
>> are 9 bugs being ignored (I noticed that all of them are unassigned).

gw...@si... said:
> No. 

I'd suggest having them turn it off or something.  You have users thinking
you're ignoring them.

>> Secondly, I've had it crash consistently when trying to use autotrace.
>> The following patch fixes (or, at least works around) the problem
>> (sc->parent is NULL):

> sc->parent should never be NULL.
> Please tell me what you do to generate the crash.

As I recall, I was doing an Autotrace using autotrace.  Consistently.
I've stopped using autotrace as it took more work to clean it up than
to do it by hand from scratch.  It's probably better on cleaner images,
but one would have to clean up the original images.

>> Third, it looks like your error handling is broken.
>> 
>> Recovering from /home/aardvark/.PfaEdit/autosave/auto007639-14.asfd... Internal Error:
>> Invalid 2nd order spline in SplineRefigure2
>> Internal Error:
>> Invalid 2nd order spline in SplineRefigure2

> Please tell me what you do to generate this error. Send me the font and the
> error recovery file (found in ~/.PfaEdit/autosave) 

Just moving points around will sometimes generate this.  Sorry, but I
already deleted the file in the autosave directory.  It was generating
NaN's and not handling them.  Ugly workaround in SolveCubicBack() follows:

--- splineutil.c.~1.56.~        2005-05-03 11:28:01.000000000 -0400
+++ splineutil.c        2005-05-21 15:20:05.000000000 -0400
@@ -399,6 +399,9 @@
     double t, val, offset;
     int first;
 
+    if ( isnan(a) | isnan(b) | isnan(c) | isnan(d) )
+      return 0.5;              /* for now... */
+
     offset=a;
     if ( a<0 ) offset = -a;
     if ( b<0 ) offset -= b; else offset += b;

BTW, it would be nice if it at least indicated which character it was
complaining about (as someone reported on sourceforge in the bugs you
are ignoring).  Better yet, indicate the offending spline.


Yet another problem I've run into is that it likes to generate corner
points on occasion when I've told it to generate curve points.  The point
menu has curve checked.  Only one control is grabbable.  Changing the
point type doesn't do anything.  And, no, the other control point wasn't
sitting on the curve point; I had two adjacent points like this where
the control points were away from the other curve point (as if I had
two corner points with a strait line between them).

-d