Re: [MiNT] [PATCH] XaAES regression in ob_fix_shortcuts()

["Helmut Karlowski" <helmut.karlowski@ish.de>]

Vincent Rivière, 27.11.2012 21:26:47:

> The first thing to do is to add an assert() to early detect a buffer
> overflow. This is easy and saves much trouble. See my assert() patch as
> attachment. It will always detect the overflow, whatever the platform is
> (not only FireTOS).
>
> assert.patch
> Detect buffer overflow in ob_fix_shortcuts(). Contributed by Vincent  
> Riviere.

I don't think it is necessary to use assert here, simply stop processing  
and emitting a message would be enough (see new patch).

> > Attached an improved version.
>
> Thanks, your patch fixes the problem :-)

I promise the buffer cannot overflow any more :-)

> Please commit.
>
> BTW, the fact that the last parameter of ob_fix_shortcuts() is a double
> pointer (void **scp) is strange. This allows to return the internally
> allocated buffer to the outside, but this functionality does not seem to  
> be
> used.

I said I'll have a look at this, but that is not urgent, the buffer-size  
is big enough atm.

> IMHO, a simple pointer and the actual array length (void *scp, long  
> scplen)
> would be much better. So the assert() will check the buffer overflow  
> against
> the real buffer size, whenever the buffer is allocated internally or  
> externally.

Currently the buffer-size is not known to ob_fix_shortcuts in that case,  
I'll probably pass it from the caller.

> Helmut, I let you commit the patches and do any additional work, if you  
> like.

Go ahead!

If you remove any line containing predef it should give no warning (I did  
not receive it ...).

So I'll commit this and your other patch or better do the  
buffer-from-caller-thing first.

--
Helmut Karlowski

Attachment: obtree.diff
Description: Binary data