urlglob does not allow dashes in IPv6 zone ID

[puckipedia]

It seems that urlglob's IPv6 parsing is less lenient than the actual URL parsing is, it doesn't allow for dashes..

I did this

marisa ~> curl -v 'http://[fe80::1%foo-bar]/`
curl: (3) bad range in URL position 9:
http://[fe80::1%foo-bar]/
        ^

I expected the following

marisa ~> curl -gv 'http://[fe80::1%foo-bar]/'
*   Trying fe80::1:80...

curl/libcurl version

curl 7.70.0 (x86_64-pc-linux-gnu) libcurl/7.70.0 OpenSSL/1.1.1g zlib/1.2.11 libssh2/1.9.0 nghttp2/1.40.0
Release-Date: 2020-04-29
Protocols: dict file ftp ftps gopher http https imap imaps pop3 pop3s rtsp scp sftp smb smbs smtp smtps telnet tftp
Features: AsynchDNS GSS-API HTTP2 HTTPS-proxy IPv6 Kerberos Largefile libz NTLM NTLM_WB SPNEGO SSL TLS-SRP UnixSockets

operating system

Linux marisa 5.6.15 #1-NixOS SMP Wed May 27 15:48:31 UTC 2020 x86_64 GNU/Linux

[dfandrich]

On Wed, Jun 17, 2020 at 10:23:14AM -0700, puckipedia wrote: marisa ~> curl -v 'http://[fe80::1%foo-bar]/` curl: (3) bad range in URL position 9: http://[fe80::1%foo-bar]/

This isn't a valid URL. The percent must be escaped, i.e. http://[fe80::1%25foo-bar]/

[puckipedia]

marisa ~> curl -v 'http://[fe80::1%25foo-bar]/'
curl: (3) bad range in URL position 9:
http://[fe80::1%25foo-bar]/

And either way, RFC6874 Section 3 suggests parsers be liberal in accepting non-percent-escaped zone IDs in IPv6 literals.

[dfandrich]

You need the -g option to use IPv6 literal addresses. This is documented somewhere.

[puckipedia]

The only reference to globbing and IPv6 I could find is https://ec.haxx.se/cmdline/cmdline-globbing, which says:

The globbing uses the reserved symbols [] and {} for this, symbols that normally cannot be part of a legal URL (except for numerical IPv6 addresses but curl handles them fine anyway). If the globbing gets in your way, disable it with -g, --globoff.

now, I can see why this is occuring, but src/tool_urlglob.c seems to imply that IPv6 addresses are not parsed as ranges by the globbing mechanism:

curl/src/tool_urlglob.c

Line 368 in 8bc25c5

/* skip over IPv6 literals and [] */

If IPv6 literal addresses in URLs shouldn't be supported without -g, why is this rule not applied consistently?

[bagder]

Look like the peek_ipv6 function needs to be fixed to deal with zone id better:

curl/src/tool_urlglob.c

Lines 322 to 350 in 8bc25c5

	static bool peek_ipv6(const char *str, size_t *skip)
	{
	/*
	* Scan for a potential IPv6 literal.
	* - Valid globs contain a hyphen and <= 1 colon.
	* - IPv6 literals contain no hyphens and >= 2 colons.
	*/
	size_t i = 0;
	size_t colons = 0;
	if(str[i++] != '[') {
	return FALSE;
	}
	for(;;) {
	const char c = str[i++];
	if(ISALNUM(c) || c == '.' || c == '%') {
	/* ok */
	}
	else if(c == ':') {
	colons++;
	}
	else if(c == ']') {
	*skip = i;
	return colons >= 2 ? TRUE : FALSE;
	}
	else {
	return FALSE;
	}
	}
	}

[bagder]

@puckipedia do you have the ability to try out my patch for this?

[puckipedia]

Looks like it works for me, thank you very much!

marisa ~> $result/bin/curl -v 'http://[fe80::1%foo-bar]/'
*   Trying fe80::1:80...
* Connected to fe80::1 (fe80::1) port 80 (#0)