Currently, argv[0] is used in regression tests and command line tools only,
and it is not in libraries.
But, I'll try to think of that.
Thanks.

On Fri, Jul 20, 2018 at 05:56:07AM -0600, Theo de Raadt wrote:
> Disagree with this.  argv[0] could be a long path, which could info
> disclose directory hierarchy into a context that bounces stderr to
> the wrong place.
> 
> Instead, the Windows compat should have a have a getprogname() stub
> which does the right thing.
> 
> Kinichiro Inoguchi <[email protected]> wrote:
> 
> > To run regress bnaddsub on Windows, I would like to supersede getprogname
> > with argv[0] since it is not on Windows.
> > 
> > OK ?
> > 
> > Index: regress/lib/libcrypto/bn/addsub/bnaddsub.c
> > ===================================================================
> > RCS file: /cvs/src/regress/lib/libcrypto/bn/addsub/bnaddsub.c,v
> > retrieving revision 1.1
> > diff -u -p -u -p -r1.1 bnaddsub.c
> > --- regress/lib/libcrypto/bn/addsub/bnaddsub.c      10 Jul 2018 16:57:50 
> > -0000      1.1
> > +++ regress/lib/libcrypto/bn/addsub/bnaddsub.c      17 Jul 2018 13:01:08 
> > -0000
> > @@ -216,7 +216,7 @@ main(int argc, char *argv[])
> >  
> >     if ((bio_err = BIO_new_fp(stderr, BIO_NOCLOSE)) == NULL) {
> >             fprintf(stderr, "%s: failed to initialize bio_err",
> > -               getprogname());
> > +               argv[0]);
> >             return 1;
> >     }
> >  
> > 

Reply via email to