Hi Mike,

that's my mess.

> In the UT_contextGlyph constructor there's a call
> 
> memcpy(&s_table[HEBREW_START], &s_table[HEBREW_END + 1] ...
> 
> As all (should) know, overlapping memcpy is undefined. If the areas
> overlap one should use memmove, but since there are quite a few
> platforms involved I didn't want to break anything by fixing this.

The gap between the overlap is 120 bytes -- I did not think there was 
a processor around that would move data around in such big chunks 
as to have a problem with this.

> Another thing is UT_contextGlyph::renderString. 28 lines into this
> function there's
> 
>   for(; j < CONTEXT_BUFF_SIZE; j++)
>     next_tmp[j] = *(next + (j + i + 1 - len));
> 
> I really don't know what this code does, and I'm not sure I want to
> know. What I do know is that the expression (next + (j + i + 1 - len))
> evaluated to an address that is not initialized (called from
> fp_TextRun::_refreshDrawBuffer). Someone knowing this code might want
> to have a look at that too.

Just having a quick look at the code, it seems to me that the 
indexing is correct; perphaps the problem is in 
fp_TextRun::_refreshDrawBuffer(). Unfortunately, I am to busy at 
the moment to be able to look into this.

Tomas

Reply via email to