Re: [musl] [PATCH v2 1/2] nftw: implement FTW_CHDIR

[Markus Wichmann <nullplan@gmx.net>]

On Sat, Jan 22, 2022 at 03:51:58PM +0100, Ismael Luceno wrote:
> @@ -133,9 +136,16 @@ int nftw(const char *path, int (*fn)(const char *, const struct stat *, int, str
>  	}
>  	memcpy(pathbuf, path, l+1);
>
> +	if (flags & FTW_CHDIR)
> +		orig_dfd = open(".", O_CLOEXEC | O_PATH);
> +
>  	pthread_setcancelstate(PTHREAD_CANCEL_DISABLE, &cs);
>  	r = do_nftw(pathbuf, fn, fd_limit, flags, NULL);
>  	pthread_setcancelstate(cs, 0);
> +	if (flags & FTW_CHDIR) {
> +		fchdir(orig_dfd);
> +		close(orig_dfd);
> +	}
>  	return r;
>  }
>
> --
> 2.33.0
>

Erm... maybe a dumb question, but what if either the open() or the
fchdir() fails? Is anything specified for that case? Is effectively
ignoring FTW_CHDIR OK in that case or would you have to signal failure?
I mean, I see that there is not much you can do in either case, but to
just fail silently seems wrong to me.

Ciao,
Markus