[BUG] Emacs thinks it has a BDF font

[BUG] Emacs thinks it has a BDF font

[David Abrahams]

In http://article.gmane.org/gmane.emacs.devel/16583/ I reported that
emacs is following an execution path which attempts to draw a BDF font
even though none is in use, resulting in a crash deep in the drawing
machinery about once a day for me.  I'm unfamiliar with this part of
Emacs' display code.  Is there someone who can give me a little
guidance about what to assert, where to set a breakpoint, or something
(anything) that will help us to track this down?

Thanks,

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com

Re: [BUG] Emacs thinks it has a BDF font

[Jason Rumney]

David Abrahams <dave@boost-consulting.com> writes:

> In http://article.gmane.org/gmane.emacs.devel/16583/ I reported that
> emacs is following an execution path which attempts to draw a BDF font
> even though none is in use, resulting in a crash deep in the drawing
> machinery about once a day for me.  I'm unfamiliar with this part of
> Emacs' display code.  Is there someone who can give me a little
> guidance about what to assert, where to set a breakpoint, or something
> (anything) that will help us to track this down?

In a previous report, you mentioned that this occured when you were
using Gnus 5.10. Does that try to use a BDF font behind your back?

If not, then you are probably looking at stack or heap
corruption. Since it is happening in display code variables that are
exected to change frequently, it will be very difficult to track
down with data breakpoints. Take a look at the structures involved,
and see if there are any arrays that might overrun. assert that the
length is within limits wherever they are written. See if you can work
out what variables are next to the font structure on the heap or stack
when the bug occurs. Add assertions to any arrays writes there too.

Re: [BUG] Emacs thinks it has a BDF font

[David Abrahams]

Jason Rumney <jasonr@gnu.org> writes:

> David Abrahams <dave@boost-consulting.com> writes:
>
>> In http://article.gmane.org/gmane.emacs.devel/16583/ I reported that
>> emacs is following an execution path which attempts to draw a BDF font
>> even though none is in use, resulting in a crash deep in the drawing
>> machinery about once a day for me.  I'm unfamiliar with this part of
>> Emacs' display code.  Is there someone who can give me a little
>> guidance about what to assert, where to set a breakpoint, or something
>> (anything) that will help us to track this down?
>
> In a previous report, you mentioned that this occured when you were
> using Gnus 5.10. Does that try to use a BDF font behind your back?

I don't think so.  I have two emacs' running at all times, only one of
which runs GNUs, and the other one is just as likely to crash
(perhaps more likely).

> If not, then you are probably looking at stack or heap
> corruption. Since it is happening in display code variables that are
> exected to change frequently, it will be very difficult to track
> down with data breakpoints. 

Or at all.

> Take a look at the structures involved,

I have, briefly.  It's hard to make much of them; they are so complex.
Is there a guide to the display structures somewhere?

> and see if there are any arrays that might overrun. assert that the
> length is within limits wherever they are written. See if you can work
> out what variables are next to the font structure on the heap or stack
> when the bug occurs. Add assertions to any arrays writes there too.

<shiver>

I guess this is going to be a tough one...

Thanks for the tips,
Dave

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com