Indeed, get_headers()'s $url has to be a valid path, because we're
passing it to php_stream_open_wrapper_ex().


 ext/standard/url.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/ext/standard/url.c b/ext/standard/url.c
index 57fd80cc1d..fe6d7f9de1 100644
--- a/ext/standard/url.c
+++ b/ext/standard/url.c
@@ -680,7 +680,7 @@ PHP_FUNCTION(get_headers)
 	php_stream_context *context;
 
 	ZEND_PARSE_PARAMETERS_START(1, 3)
-		Z_PARAM_STRING(url, url_len)
+		Z_PARAM_PATH(url, url_len)
 		Z_PARAM_OPTIONAL
 		Z_PARAM_LONG(format)
 		Z_PARAM_RESOURCE_EX(zcontext, 1, 0)


I don't see the need for an regression test (ZPP only).

Note: the code in the test script is in no way secure. parse_url() does not guarantee the resulting URL is valid (or even HTTP URL) - it just tries its best to parse it in assumption it's valid. If you need URL validation (as you should if you importing external URLs) other methods should be used. 

That said. no reason for get_headers() to accept URLs with nulls, so it shouldn't.

@stas That's true. I may have oversimplified the example. My main point to get across was that even if validation exists, it wouldn't necessarily check the subdomain using a whitelist. There are too many characters valid in subdomains now:
https://stackoverflow.com/questions/7111881/what-are-the-allowed-characters-in-a-subdomain/22319900#22319900