Re: Bug#622365: Fix FTBFS of tofrodos for GNU/Hurd
Svante Signell, le Tue 12 Apr 2011 21:46:09 +0200, a écrit :
> On Tue, 2011-04-12 at 20:56 +0200, Samuel Thibault wrote:
> > Svante Signell, le Tue 12 Apr 2011 19:27:10 +0200, a écrit :
> > > return -1 ;
> > > }
> > > /* If we reach here, "filename" is not a symbolic link */
> > And thus we didn't call realpath at all. I believe the free below
> > shouldn't be done, it's probably only by chance that it doesn't segfault
> > in your test (by being NULL by chance).
> Yes, we did not call realpath but the variable is still declared but not
> malloced. Does that mean we shouldn't free() it?
> > > +#ifdef __GNU__
> > > + free (realfilepath);
> > > +#endif
> > > return openandconvert( filename );
> > > }
> OK, thanks for your comments. The biggest problem with malloced
> storage, is where to put the free() statements. To do that you need to
> follow the program logic and find out every statement with a possible
> exit. This can be very difficult sometimes, at least I think so.
> Final version??
Seems good to me.