fVDI TTF rendering (was: RE: [Aranym-user] X Server and fVDI)

Johan Klockars johan at klockars.net
Thu Feb 23 03:00:15 CET 2006


>I find it very unlikely that there is a 'fault' in my system.
>  
>

I'm not saying that there's anything wrong with your system.
After all, mine's the one that keeps crashing. ;-I

>I just made sure I updated all sources and compiled the binaries
>again. I have tested the TOS4.04/fVDI and
>EmuTOS/MiNT/fVDI-started-after-mint
>setups and both work with no obvious problem with or without
>antialiasing.
>  
>

When I tested CVS ARAnyM with plain TOS 4.04+fVDI, PapyrusX seemed fine 
without anti-aliasing. With anti-aliasing turned on, PapyrusX reported a 
bus error after drawing about half the first page of the large test 
document.
With MiNT+XaAES it doesn't get that far, with or without anti-aliasing. 
That stack overflow dialog I mentioned appears and the application exists.
MiNT+MyAES, which I just tried, does however (seem to) work fine. Even 
with anti-aliasing!

It feels reasonable to assume that this must be some kind of memory 
overwriting bug somewhere in the fVDI code. Possibly XaAES detects it by 
checking the application stack, which could explain that "stack 
overflow" dialog and exit.
It's really strange that an older ARAnyM build can make things work 
better (or at least seem to), though.

>BTW: There are a bunch of strcpy() used in the new functions
>that copy strings to a limited space (strncpy() should be used there
>IMHO).
>  
>

You're quite right. Programming in the middle of the night isn't always 
the best idea. ;-)


Anyway, the latest fVDI version now has significantly faster internal 
rendering of cached vector character bitmaps. I'm not sure if this will 
really be noticeable under ARAnyM without some not yet existing 
benchmark program, however.
(If anyone happens to know about any benchmarks that test GDOS fonts at 
all, and vector fonts in particular, I'd be interested in hearing about 
them.)
More importantly, it fixes a vector font rendering problem I had never 
seen before I was debugging these new speedups (I introduced this bug 
the last time I was working on the very same code. ;-). If a character 
partially overlapped the next one, it could sometimes (particularly at 
large sizes) be displayed incorrectly.

-- 
| Why are these | johan at klockars.net
|  .signatures  |
| so hard to do | http://www.klockars.net
|     well?     | (fVDI, MGIFv5, QLem)



More information about the cz-bobek-lists-aranym-user mailing list