[9fans] stringwidth leak or ignorance?

[9fans] stringwidth leak or ignorance?

[Fco. J. Ballesteros]

Hi,

	I've seen that  at point 
sys/src/libdraw/stringwidth.c:46
  45		if(subfontname){
>46			if(_getsubfont(f->display, subfontname) == 0){
 47				def = f->display->defaultfont;
 48				if(def && f!=def)
 49					f = def;
 50				else
 51					break;

there seems to be a leak.
_getsubfont returns a Subfont, which is not freed. However, there
seems to be some side effect, because if I free the subfont, the routine breaks.

This came from using leak on a program using stringwidth a lot. The program
shows leaks that come from this point. So I tried to fix it. I still don't know enough
of font handling to propose a course of action. Any suggestion?

If I come up with a fix, I'll send a patch anyway.

thanks

Re: [9fans] stringwidth leak or ignorance?

[Rob Pike]

the call to _getsubfont just puts the subfont into the subfont
cache so the character can be looked up there the next time
around the loop.  there should be no leak.  freeing the subfont
will break the cache.

-rob

On 5/26/05, Fco. J. Ballesteros <nemo@lsub.org> wrote:
> 
> Hi,
> 
>         I've seen that  at point
> sys/src/libdraw/stringwidth.c:46
>   45            if(subfontname){
> >46                     if(_getsubfont(f->display, subfontname) == 0){
>  47                             def = f->display->defaultfont;
>  48                             if(def && f!=def)
>  49                                     f = def;
>  50                             else
>  51                                     break;
> 
> there seems to be a leak.
> _getsubfont returns a Subfont, which is not freed. However, there
> seems to be some side effect, because if I free the subfont, the routine breaks.
> 
> This came from using leak on a program using stringwidth a lot. The program
> shows leaks that come from this point. So I tried to fix it. I still don't know enough
> of font handling to propose a course of action. Any suggestion?
> 
> If I come up with a fix, I'll send a patch anyway.
> 
> thanks
> 
>