[imp] Login via imp successful but list mboxes leads to immediate logout

Herbert Pophal pophal at tubit.tu-berlin.de
Wed Feb 10 14:24:44 UTC 2016



On 28.01.2016 16:58, Michael J Rubinsky wrote:
>
> Quoting Herbert Pophal <pophal at tubit.tu-berlin.de>:
>
>> On 13.11.2015 16:18, Michael J Rubinsky wrote:
>>>
>>> Quoting Jan Schneider <jan at horde.org>:
>>>
>>>> Zitat von Herbert Pophal <pophal at tubit.tu-berlin.de>:
>>>>
>>>>> On 11.11.2015 17:46, Michael J Rubinsky wrote:
>>>>>>
>>>>>> Quoting Herbert Pophal <pophal at tubit.tu-berlin.de>:
>>>>>>
>>>>>>> Webmail Edition 5.1.3
>>>>>>> IMP 6.2.2
>>>>>>> PHP 5.5.24/5.6.8 (fcgid connected)
>>>>>>> imap server: RHEL 6.6
>>>>>>> cyrus-imapd-2.3.16-6.el6_2.5
>>>>>
>>>>> NOPE: 2.3.16-13.el6_6, Sorry
>> ...
>>>>>>> CACHE: Using the Horde_Imap_Client_Cache_Backend_Cache storage
>>>>>>> driver.
>>>>> C: 3 NAMESPACE
>>>>> S: * NAMESPACE (("" "/")) (("Other Users/" "/")) (("Shared Folders/"
>>>>> "/"))
>>>>> S: 3 OK Completed
>>>>>>> Command 3 took 0.005 seconds.
>>>>> C: 4 LSUB "" INBOX
>>>>> C: 5 LSUB "" *
>>>>> C: 6 LSUB "" "Other Users/*"""
>>>>> C: 7 LSUB "" "Shared Folders/*"""
>>>>
>>>> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>>>> This is the issue.
>>>> I remember seeing this recently, please check the mailing list archive
>>>> and bug tracker.
>>>
>>>
>>>
>>> https://bugs.horde.org/ticket/13825
>>>
>>
>> I feel rather sorry not having responded this more recently.
>>
>> Thank you very much.
>>
>> AFAICS this is not fixed in PHP 5.6, neither upstream nor at
>> iuscommunity. I upgraded a test installation to PHP 5.5.20. Did you
>> work around this bug somewhere in Horde in order we can probably
>> upgrade to a PHP 5.6 aware Horde version? Or what else do you recommend?
>
> No, this is not worked around in Horde since the change is a BC break.
> Changing to support the broken behavior in PHP would break installations
> using versions of PHP with the non-broken behavior.
>
But prevents from upgrading to php > 5.5.20/5.6.4.

I repatched a php 5.6.17 with your pull request. Due to the (implicit) 
--fuzz=0 argument to patch when building an rpm, commit d263fea failed 
since line 599 reads

php_stream_bucket_append(brig_inp, bucket TSRMLS_CC);

instead of

php_stream_bucket_append(brig_inp, bucket);

as expected by the commit.


As not being familiar with what's going on under the hood, what are the 
consequences to Horde when applying/not applying both commits 0292cb8 
and d263fea, but having reverted to the old feof() bvehavior?

Herbert


More information about the imp mailing list