Subject: Re: Segfault in libpspell
From: Kevin Atkinson (kevina@users.sourceforge.net)
Date: Wed Nov 15 2000 - 00:36:23 CST
On Wed, 15 Nov 2000, Sam TH wrote:
> On Wed, Nov 15, 2000 at 01:04:07AM -0500, Kevin Atkinson wrote:
> > On Tue, 14 Nov 2000, Sam TH wrote:
> >
> > > On Tue, Nov 14, 2000 at 07:29:43PM -0500, Kevin Atkinson wrote:
> > > > On Tue, 14 Nov 2000, Matt Brubeck wrote:
> > > >
> > > > > Also, pspell_manager_check() should not segfault when passed a null
> > > > > pointer as its first argument. This is a bug in Pspell.
> > > >
> > > > Um, Why? You are violating an implied precondition of passing in a valid
> > > > pointer.
> > > >
> > >
> > > Because programs shouldn't segfault. This is certainly our fault,
> > > for passing the NULL pointer. However, it is preferable to check
> > > that, than to just segfault.
> >
> > A library can certainly segfault. The behavior of many libc functions
> > are undefined if you give them invalid input. Similarly the behavior of
> > many pspell functions are undefined if you give them invalid input.
> >
> > Just what would you like me to do if I get a null pointer?
>
> Well, as far as I can tell from Dom's code, pspell_manager_check() returns
> 1 if found, 0 if not found, and -1 on error. Therefore, if it was passed
> a null pointer, it could return -1. The following patch does just that
> (actually it checks both arguments, which should be the right thing to do):
1) If it returns -1 it should also set the appropriate error message.
2) The behavior of pspell_manager_check should be identical to the
behavior of m->check();
3) It is unnecessary in my view as the behavior is undefined if you pass it in
invalid input.
Thus your patch is rejected. As I said before I do NOT think this is a
bug nor will I ever consider undefined behavior due to invalid input a
bug. Sometimes I will check if the input is valid with an assert however
in the case of null pointers this is unnecessary because dereferencing a
null pointer guaranties a segfault. A backtrace of where the program
segfaulted will clearly show a null pointer being passed in.
I am sorry I do not agree with you here.
---
Kevin Atkinson
kevina at users sourceforge net
http://metalab.unc.edu/kevina/
This archive was generated by hypermail 2b25 : Wed Nov 15 2000 - 00:26:35 CST