Should be able to spot the reverse overlap. The memcpy vs memmove difference is pretty simple, and has not false positived once.
Claudio Jeker <cje...@diehard.n-r-g.com> wrote: > On Mon, Dec 30, 2019 at 10:31:30AM +0100, Denis Fondras wrote: > > On Sun, Dec 29, 2019 at 11:56:57PM +0100, Theo Buehler wrote: > > > > That's a problem on my side, I will sort it and retry. > > > > > > This diff fixes it: > > > > > > Index: libvips/iofuncs/init.c > > > --- libvips/iofuncs/init.c.orig > > > +++ libvips/iofuncs/init.c > > > @@ -858,7 +858,7 @@ extract_prefix( const char *dir, const char *name ) > > > for( i = 0; i < (int) strlen( vname ); i++ ) > > > if( vips_isprefix( G_DIR_SEPARATOR_S "." G_DIR_SEPARATOR_S, > > > vname + i ) ) > > > - memcpy( vname + i, vname + i + 2, > > > + memmove( vname + i, vname + i + 2, > > > strlen( vname + i + 2 ) + 1 ); > > > if( vips_ispostfix( vname, G_DIR_SEPARATOR_S "." ) ) > > > vname[strlen( vname ) - 2] = '\0'; > > > > > > > Thank you Theo. > > Any idea why sometime it would break and sometime it would work ? > > If strlen( vname + i + 2 ) < 2 then there is no overlap. Not sure if that > can explain why it works sometimes but fails other times. > > -- > :wq Claudio >
