Re: [ft-devel] endless loops in bytecode

[Werner LEMBERG]

>> My only idea up to date is to count JMPR, JROT, and JROF separately,
>> using a threshold of, say, 10000.
> 
> I would run a total jump count of any kind though and stop after
> 1000 jumps of any kind.

Exactly.  However, I will probably also consider the LOOPCALL counter.

> Enough is enough.  The bytecode length is limited to 65536 (no?).

This is irrelevant to the problem – due to repeated calls of bytecode
functions I can imagine that for complicated glyphs a 2^16 limit gets
easily overflowed.

> In the worst case scenario, 1000 jumps will produce 65 million
> instructions.

We already have a counter that limits the number of executed
instructions per glyph to 1000000.

Another idea was to switch off hinting completely for a font in case
we ever hit the bytecode overflow limit.  This I probably won't do,
though.


    Werner