Re: [PATCH] glob: fix / matching

[Rich Felker <dalias@libc.org>]

On Sat, May 27, 2017 at 09:59:18PM -0400, Rich Felker wrote:
> On Fri, May 19, 2017 at 04:59:13PM +0200, Julien Ramseier wrote:
> > glob(3) currently fails matching "/", due to any '/' being
> > removed from the start of the pattern before checking if
> > it's empty.
> > 
> 
> > diff --git a/src/regex/glob.c b/src/regex/glob.c
> > index 5b6ff12..53f3a4e 100644
> > --- a/src/regex/glob.c
> > +++ b/src/regex/glob.c
> > @@ -156,18 +156,11 @@ static int sort(const void *a, const void *b)
> >  
> >  int glob(const char *restrict pat, int flags, int (*errfunc)(const char *path, int err), glob_t *restrict g)
> >  {
> > -	const char *p=pat, *d;
> > +	const char *p = pat;
> >  	struct match head = { .next = NULL }, *tail = &head;
> >  	size_t cnt, i;
> >  	size_t offs = (flags & GLOB_DOOFFS) ? g->gl_offs : 0;
> >  	int error = 0;
> > -	
> > -	if (*p == '/') {
> > -		for (; *p == '/'; p++);
> > -		d = "/";
> > -	} else {
> > -		d = "";
> > -	}
> >  
> >  	if (!errfunc) errfunc = ignore_err;
> >  
> > @@ -179,12 +172,19 @@ int glob(const char *restrict pat, int flags, int (*errfunc)(const char *path, i
> >  
> >  	if (strnlen(p, PATH_MAX+1) > PATH_MAX) return GLOB_NOSPACE;
> >  
> > -	if (*p) error = match_in_dir(d, p, flags, errfunc, &tail);
> > +	if (*p) {
> > +		const char *d = "";
> > +		if (*p == '/') {
> > +			for (; *p == '/'; p++);
> > +			d = "/";
> > +		}
> > +		error = match_in_dir(d, p, flags, errfunc, &tail);
> > +	}
> 
> I'm confused how this patch differs from just removing the "if (*p)"
> condition before calling match_in_dir. Does match_in_dir actually work
> if p points to an empty string? I thought not...

Hmm, just removing the "if (*p)" seems to make it work; it looks like
match_in_dir covers this case fine. I'd like a second opinion on
whether this looks okay since I haven't touched this code in years,
but I suspect it'll turn out to be an okay fix for now...

> > Also, any feedback concerning
> > http://www.openwall.com/lists/musl/2017/01/12/5 ? 
> 
> I'll have to actually trace through what it does to figure out why
> that's happening; the intent was that it work for all components, I
> think..
> 
> Really the whole glob() implementation could use a complete rewrite.

...but I still think the code is due for a rewrite. Just the fact that
it collapses all ///// to / is rather bad.

Rich