Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:54600
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: error (pb1.pair.com: domain lerdorf.com from 209.85.210.172 cause and error)
Sender: Rasmus Lerdorf <rasmus@lerdorf.com>
Message-ID: <4E4843B0.70605@php.net>
Date: Sun, 14 Aug 2011 14:52:48 -0700
Organization: PHP Development Team
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:5.0) Gecko/20110627 Thunderbird/5.0
MIME-Version: 1.0
To: Stas Malyshev <smalyshev@sugarcrm.com>
CC: Derick Rethans <derick@php.net>, 
 PHP Internals <internals@lists.php.net>
References: <CABBJUpfez0_CKkiFSVVmYfxbF7H30Av2y9D1Nww6PE5v6dKPUw@mail.gmail.com> <CAM4sM9Hxn=2FJ4d51ozbKuGM4e3BMMUQ38WsEs27WAu919XSjg@mail.gmail.com> <CABBJUpdONFHHYpdr4RvxGRyJ13M9yONVzb-aSjx33QKziALkLA@mail.gmail.com> <CABBJUpeoDnuT7eXC6UzUFQni0V0zH_mivHiX=3j=UjCKJK-z7Q@mail.gmail.com> <CABBJUpdXUOjyWpXeRovNaFgDMQHfM5MkXXto+Rybej2u+W4xFw@mail.gmail.com> <alpine.DEB.2.02.1108141219470.4043@whisky> <CAH-PCH6_sTBGqCrCp-pgdi3DJfFX-3ZsvSr8ibbrOtKQP0qpTA@mail.gmail.com> <alpine.DEB.2.02.1108141359340.4043@whisky> <4E48121A.3090007@sugarcrm.com> <4E48134E.4030708@sugarcrm.com> <4E481695.6040900@php.net> <4E483804.7070009@sugarcrm.com> <4E484037.2080107@php.net>
In-Reply-To: <4E484037.2080107@php.net>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Subject: Re: [PHP-DEV] [VOTE]strn(case)cmp supporting a negative length as
 its third paramter
From: rasmus@php.net (Rasmus Lerdorf)

On 08/14/2011 02:37 PM, Rasmus Lerdorf wrote:
> On 08/14/2011 02:03 PM, Stas Malyshev wrote:
>> Hi!
>>
>> On 8/14/11 11:40 AM, Rasmus Lerdorf wrote:
>>> My main issue with changing strncmp/strncasecmp is that these are
>>> currently exact mappings of the underlying libc functions. For people
>>
>> And why should anybody care? 99% of people using PHP never used a libc
>> function and can hardly tell libc from gcc. If we can extend this
>> function with useful functionality, nobody cares about what libc does.
>>
>>> For example, I could imagine people writing code along these lines:
>>>
>>> $len = strlen($user_data) - strlen($suffix);
>>> if(!strncmp($user_data, $string, $len)) {
>>>     // do something
>>> }
>>
>> Warning doesn't fix the bug - and unless you're in 0.0001% of the
>> population that actually reads the logs daily and checks every message
>> there it would be little to help you. We should have more useful
>> functions, not more warnings. Warning won't make this code to work.
> 
> I agree, however this change would potentially change the return value
> of the function. Before it would warn and not match. Even if you never
> saw the warning, at least length -1 would not give you a match. Now if
> the user data happens to end with the right character we now have a
> string match which is not at all what the code was written to do.

Put more succinctly. Subtle BC breaks like this worry me. Any strncmp()
call with a computed length where that length may in some cases go
negative will now potentially return a match where it wouldn't before.
This would be very hard to track down. And the reason for introducing
this subtle BC break is so that you can rewrite:

   if (substr("prefix_num", -3) == "num") {
        echo "they have same suffix\n";
   }


into:

   if (strncmp("prefix_num", "num", -3) === 0) {
        echo "they have same suffix\n";
   }

That doesn't seem like a big win to me.

-Rasmus