Re: OpenType with CFF outlines

[Tom Kacvinsky]

More information:  the only time that the name_index data gets stomped on
(that I can tell) is on a Linux box.  DEC Unix and Solaris don't exhibit
this problem.  Strange...

On Tue, 27 Jun 2000, Tom Kacvinsky wrote:

> I can tell you one place where memeory is getting stomped on:
> the name_index data.
> 
> Disassembling the font one way, I see that the name data off_size
> is 1, there is exactly one entry and the two offsets are 1 and 21,
> resp.  When T2_Get_Name is called, the first offset is all of a
> sudden 408.  Looking at the intervening calls, I can't see where
> memory is getting stomped, unless you have change t2parse.c
> *significantly*.
> 
> On Tue, 27 Jun 2000, David Turner wrote:
> 
> > Sorry for taking so long to answer, but I believe that the problem
> > with CFF comes from a bug in the loader that much probably overwrites
> > some memory structures (electric fence doesn't catch everything, as
> > you probably know)..
> > 
> > I'm currently hacking this code (in order to support CID fonts) and
> > quite a few things have changed. Please wait for a few days for more
> > info..
> > 
> > Cheers,
> > 
> > - David
> > 
> > Werner LEMBERG a écrit :
> > > 
> > > > I can reproduce this problem on the following platform:
> > > >
> > > > Linux trillian 2.2.15 #1 Tue May 16 09:58:41 EDT 2000 i686 unknown
> > > >
> > > > freetype was compiled with egcs-2.91.66.  Here is the bt (looks
> > > > similar to Sivan's):
> > > 
> > > Can you try to link the demo programs with efence (or a similar memory
> > > alloc debugger library)?  Maybe this yields more details.
> > > 
> > >   make EFENCE=-lefence
> > > 
> > >       Werner
> > 
>