[php-maint] Bug#572601: Bug#572601: Bug#572601: Bug#572561: php5: crypt() output inconsistency between version 5.2 and 5.3

[Ondřej Surý]

forwarded 572601 http://bugs.php.net/bug.php?id=51247
thank you

On Tue, Mar 9, 2010 at 17:00, Raphael Geissert <geissert at debian.org> wrote:
> 2010/3/9 Ondřej Surý <ondrej at debian.org>:
>> I just checked ext/standard/config.m4 in 5.3.2 and it contains this code:
>>
>> if test "$ac_cv_crypt_blowfish" = "no" || test "$ac_cv_crypt_des" =
>> "no" || test "$ac_cv_crypt_ext_des" = "no" || test "x$php_crypt_r" =
>> "x0"; then
>>
>> And since:
>>
>> checking for crypt in -lcrypt... (cached) yes
>> checking for standard DES crypt... yes
>> checking for extended DES crypt... no
>> checking for MD5 crypt... yes
>> checking for Blowfish crypt... no
>> checking for SHA512 crypt... no
>> checking for SHA256 crypt... no
>>
>> BTW, SHA512 and SHA256 test are just broken.
>>
>
> Have you reported them upstream?

Yes, I did. Guess what, they recommend not to use system libraries.
What a surprise.

>> The reason why it fails is that from php 5.3.0 there is a new
>> "portable" re-implementation of crypt function which is used when any
>> of supported crypt function is not found.
>
> Yes, so I don't think we should simply disable it because we wouldn't
> be providing some of the features that should now be taken for granted
> on any platform.
> I would prefer if our glibc provided them.

I'll have a patch ready this evening (if everything works fine)..

BTW... am I correct that RETVAL_STRING doesn't exit the function?
(Looks like the case according to zend_API.h)

Ondrej
-- 
Ondřej Surý <ondrej at sury.org>
http://blog.rfc1925.org/