php.net |  support |  documentation |  report a bug |  advanced search |  search howto |  statistics |  random bug |  login
Request #51073 crc32() function result differs on 64-bit Linux platforms and others
Submitted: 2010-02-17 17:47 UTC Modified: 2010-02-19 13:52 UTC
From: jian at theorchard dot com Assigned:
Status: Not a bug Package: Feature/Change Request
PHP Version: 5.3.1 OS: Linux RHEL 5
Private report: No CVE-ID: None
View Developer Edit
Welcome! If you don't have a Git account, you can't do anything here.
If you reported this bug, you can edit this bug over here.
(description)
Block user comment
Status: Assign to:
Package:
Bug Type:
Summary:
From: jian at theorchard dot com
New email:
PHP Version: OS:

 

 [2010-02-17 17:47 UTC] jian at theorchard dot com
Description:
------------
According to bug# 36306, this issue should have been fixed.  But I found it not entirely fixed across all platforms.

crc32 is returning an unsigned integer on a 64-bit Linux platform.  It does return the signed integer from a 64-bit Windows platform as well as 32-bit Windows/Linux platforms.

I found PHP_INT_SIZE and PHP_INT_MAX constants have different values on that 64-bit Linux server with values 8 and 9223372036854775807 respectively.  They are 4 and 2147483647 on other 32-bit platforms and 64-bit Windows platform.

Reproduce code:
---------------
<? echo crc32("884385799717_1_1") . "\n"; ?>

Expected result:
----------------
I expect to see value -676770709 both on 32-bit and 64-bit platforms.

Actual result:
--------------
On 64-bit Linux platform I see 3618196587
On 64-bit Windows platform I see -676770709
On 32-bit platforms I see -676770709

Patches

Pull Requests

History

AllCommentsChangesGit/SVN commitsRelated reports
 [2010-02-17 18:21 UTC] pajoye@php.net
It is the signed output of an unsigned value. The values are correct (there was other bug reports about this, check them to get a more verbose explanation).
 [2010-02-17 22:23 UTC] jian at theorchard dot com
I certainly understand that it is the signed output of an unsigned value.  But when crc32 function is used for checksum purpose, I'm expecting it to return the same checksum value from under different platforms.  

Apparently, this unsigned representation of a signed integer as the result of the function call has caused a lot of checksum failures in our applications.

I would like this behavior to be fixed in upcoming releases.

I wouldn't have created this bug report if I could find my solution or existing reports that you already have.  If you can provide me with the bug numbers of the other bug reports you mentioned, I'd be more than happy to study them.

Thank you very much.
 [2010-02-17 22:32 UTC] derick@php.net
Instead of a plain echo, you can also do:

printf("%u", crc32("884385799717_1_1"));

that should output the same string on every platform (that is at least 32bit).

The reason why it is different on Windows vs. (any other) Unix, is because on Windows even on 64 bit CPUs, "long" (the data type we use for integers) is still only 32bit; on Unices, the "long" type is usually 64 bit on 64 bit CPUs.
 [2010-02-17 22:38 UTC] jian at theorchard dot com
Thanks Derick.  That was much more helpful.  Unfortunately, I cannot use unsigned integer as our checksum.  It has to be signed.  I do in fact have a work around function built to solve this issue.  I was just wondering whether this will get fixed in the future or not.  Since it's a platform dependent issue, it most likely won't get fixed I think.

Below is the function I have posted against the crc32 function as a note.  I hope this can help others out.  Thanks again.

function get_signed_int($in) {
    $int_max = pow(2, 31)-1;
    if ($in > $int_max){
        $out = $in - $int_max * 2 - 2;
    }
    else {
        $out = $in;
    }
    return $out;
}
 [2010-02-18 08:26 UTC] jian at theorchard dot com
On a second thought, however, this still poses an issue.  As a programmer, I shouldn't be worreid about all these reformatting and work-arounds.  If my function returns me a chicken under development environment, it shouldn't return me an egg once it's released to the production environment.  

Even though it's the same value, they should also be represented in the same way regardless which platform this function is used.

This issue should still be classified as a valid bug in my opinion.
 [2010-02-18 08:34 UTC] rasmus@php.net
It does return the correct set of bits.  Anything we do to change what 
is currently being returned will seriously confuse everyone and the 
documentation is quite clear on this.
 [2010-02-19 04:20 UTC] jian at theorchard dot com
All I was saying is to point out that crc32 function is unable to return the same result represented in the same way under any different platforms.

The documentation does say to use sprintf function to format the result to be an unsigned integer in order to the get same expected result from all platforms.

In the real world, we actually do need to use signed checksum in some cases.  So with this shortcoming in mind, crc32 function cannot achieve a truly signed checksum purpose.
 [2010-02-19 05:15 UTC] rasmus@php.net
Why can't you always use a signed checksum if that is what you want?  
crc32 simply returns 32 bits.  You always get the same 32 bits 
regardless of the platform.  If you want to always represent that as a 
signed 32-bit integer, then do so.  The fact that some platforms have 
the ability to show those 32 bits without messing up the sign is 
irrelevant.  Just emulate it with something stupidly simple like:

$max32 = 2147483648;
$crc = crc32("884385799717_1_1");
if($crc>$max32-1) {
  $crc = $crc-2*$max32;
}
echo $crc;
 [2010-02-19 13:33 UTC] jian at theorchard dot com
Did you not read my previous post?  That's the same exactly function I have already posted.  Yes, it is very stupid to have to do that on the user's end.
 [2010-02-19 13:36 UTC] pajoye@php.net
Not too hard to understand that PHP has no unsigned integer type.
 [2010-02-19 13:39 UTC] rasmus@php.net
But, like I said there is nothing to do here.  We have to return the 
exact bits.  Anything we do other than that will cause all sorts of 
chaos and confusion.  So I don't understand what you are asking for,  A 
new datatype in PHP for this?  That's not going to happen obviously.
 [2010-02-19 13:52 UTC] jian at theorchard dot com
As I had previously said, forget it because it's platform related 
issue not PHP's fault.  I was trying to say it would be nice if the 
end user didn't need to do this kind of re-formatting.  It would 
nice the function acts as a black box.  That's all.  Thank you all 
for replying to my post.  And I feel very sorry to have wasted all 
of you time on this stupidity.  At least my boss was very 
understanding what I had to deal with.  Please close this ticket.  
I'm not going to post anything more here.  Thank you.
 
PHP Copyright © 2001-2024 The PHP Group
All rights reserved.
Last updated: Sun Dec 22 06:01:30 2024 UTC