[GRASS-dev] [grass-code I][455] r.proj segfaults after Jul 14 changes

[Brad Douglas]

On Thu, 2007-08-09 at 09:27 -0500, William Kyngesburye wrote:
> On Aug 9, 2007, at 1:57 AM, Brad Douglas wrote:
> 
> >> Not to mention the other modules I found null messages in, as I
> >> mentioned in a followup to the bug report.
> >>
> >> If (NULL) is indeed improper use of the message functions, I could
> >> take care of them - I have my own CVS copy of r.proj.seg updated
> >> already, and I found those others already.
> >
> > That is blatantly wrong.  Does anyone actually test code they write?
> > Please commit your changes.  I'll update the documentation.
> >
> That's interesting, because when I looked at the r.proj main.c  
> history, the "Message/Debug cleaning" revision explicitly changed a  
> couple G_message("") to G_message(NULL) (among other changes).

This is cause for some concern.  If unsure, glance at the function code.
It would have been more than obvious.

> To be fair, it may be that it worked for them (it only seems to cause  
> trouble on OSX).  They may or may not *know* if it's wrong (I didn't,  
> tho I understood what it was meant to do).

Not all NULLs are treated equally by all systems.  Turning -Wall on
should have revealed it, too.  It may have been an honest mistake, but
it was a careless mistake.

> Anyways, I've changed all the NULL messages I found in CVS now.

Thank you.  I updated the documentation this morning to explicitly warn
against it.


-- 
Brad Douglas <rez touchofmadness com>                    KB8UYR/6
Address: 37.493,-121.924 / WGS84    National Map Corps #TNMC-3785