[blfs-dev] PHP-5.5.10 error: 'tsrm_ls' undeclared [Was: ... #4813: SWIG 3.0.0]

Pierre Labastie pierre.labastie at neuf.fr
Wed Mar 26 13:53:20 PDT 2014


Le 22/03/2014 18:57, Pierre Labastie a écrit :
> Le 22/03/2014 18:17, Fernando de Oliveira a écrit :
>> Em 22-03-2014 13:20, Pierre Labastie escreveu:
>>> Le 22/03/2014 15:59, Fernando de Oliveira a écrit :
>>>
>>>> Now, it would be interesting hearing from Pierre.
>>>>
>>>> This is the sed I am intending to include in PHP:
>>>>
>>>>
>>>>       sed '/APACHE_THREADED_MPM=/d' \
>>>>           -i sapi/apache2handler/config.m4 \
>>>>           -i configure &&
>>>>
>>>>
>>>> Comment in ArchLinux:
>>>> "	# Just because our Apache 2.4 is configured with a threaded MPM
>>>>           by default does not mean we want to build a ZTS PHP.
>>>> 	# Let's supress this behaviour and build a SAPI that works fine
>>>>            with the prefork MPM."
>>>> In BLFS:
>>>>
>>>> Command Explanations:
>>>>
>>>> sed '/APACHE_THREADED_MPM=/d' ... build SAPI, instead ZTS PHP
>>>>
>>>> If someone objects, please tell here.
>>>>
>>>> If someone could help me with the "Command Explanations:", I would much
>>>> appreciate. Thanks.
>>>>
>>>> Anyway, will wait Pierre, before committing.
>>>>
>>>
>>> Back home after my hair cut... Seems you guy came to a solution while I was off.
>>> When I saw the error this morning, I remembered seeing something about tsrm_ls
>>> before, but could not remember where, sorry Fernando.
>>>
>>
>>
>>> - test if the sed allows the swig tests to pass.
>>
>> When I wrote that I wanted to hear from you was just this: if the sed
>> solves the test problem.
> 
> Well, I am sorry to tell that the tests do not pass after building php with
> the sed. There is also this thread:
> http://sourceforge.net/p/swig/mailman/message/31802171/
> 
> where a test case showing the exact same bug is reported, but it is said the
> bug is fixed in 3.0.0
> 
> Maybe it is just not...


Actually, I think _there is_ a bug in SWIG.

With php as in the book (i.e, sapi is apache2handler, or configure switch is
--with-apxs2):
applying this sed to swig, before configure, allows those tests to pass:
------------------------
sed -i '/t_ouput_helper.*zval/a\ \ TSRMLS_FETCH();' Lib/php/utils.i
------------------------

I have reported upstream, and am waiting for an answer.

There are still segfaults in the test, though. I wonder whether they have some
relation with the segfaults during PHP configure.

Pierre



More information about the blfs-dev mailing list