Jani Taskinen wrote:

> Why not make it standalone extension?
> Just check the ext/msession/ extension which has done
> this same thing.
> 
> --Jani


Then msession may have the same problem that I mentioned.
Since session module can be disabled, if users load
extension depands on session module, it will end up with
undefined symbol error. This is not nice...

This is expected when user compile CGI verion with
--disable-session and compile with other SAPI without
--disable-session. (There is work around as both of us
already knows, though)

Since extension for session module is rather simpile one,
it can be avoided by defining globals for all builds.
However, it may not be the case for yet known new modules
that have child modules.(Even with smart installer, there
will be users notice this problem.)

We can live with undefined symbol error, but we know
users will submit bug reports for that. :)

My opinion is, if module has parent module, child module
should be compiled with parent module unless the child
module is used not often. I think pgsql session handler
will be used by a lots of users. I hope ;)

I hope everyone agrees on this, since I don't want to
reply all bug report like "Hey, I got undefined symbol
error! Why?" for all modules that has child modules...

--
Yasuo Ohgaki


> 
> On Thu, 20 Dec 2001, Yasuo Ohgaki wrote:
> 
> 
>>Yasuo Ohgaki wrote:
>>
>>
>>>Martin Jansen wrote:
>>>
>>>
>>>>On Thu, 20 Dec 2001 14:41:43 +0900, Yasuo Ohgaki wrote:
>>>>
>>>>
>>>>
>>>>>4) Where document should be?
>>>>>
>>>>>
>>>>If your session handler will be part of PECL, it should be documented
>>>>in the upcoming PEAR manual. Have a look at the CVS module peardoc
>>>>to get a feeling for it.
>>>>
>>>
>>>
>>>Thank you for your reply.
>>>It may be OK to put document for modules that does not have
>>>function at all and does not work with certain modules is not
>>>compiled in. :) Any objection for this?
>>>
>>>We can live with "undefined symbol" errors also.
>>>However, I really don't want "undefined symbol" error because
>>>I'm sure users will submit bug reports for this. Besides, it's
>>>not the right way to manage software. IMHO.
>>>
>>
>>I'll try be more clear on this.
>>Do you want CGI version to fail to run just because GCI version
>>does not have session module?
>>(i.e. session save handler is loaded in php.ini for other SAPI,
>>and there is line for loading session save handler, or modules
>>for a module)
>>
>>--
>>Yasuo Ohgaki
>>
>>
>>
>>>>Basically I think it would be nice to have this in PECL, since it
>>>>is no real 'core' part of PHP and so it does not really fit in
>>>>php4/ext, eventhough it requires the session extension to be
>>>>enabled.
>>>>
>>>>
>>>I should explicitly have written about technical issues.
>>>You missed issues here....
>>>
>>>1) It highly depends on Session module.
>>>2) It does not work as standalone module at all.
>>>3) It will fail to load with undefined symbol when PHP is
>>>   compiled with --disable-session.
>>>3) It does not provide any function to users.
>>>4) Where document should be?
>>>5) How to handle include files required for session save
>>>   handler module?
>>>
>>>Session save handler modules are not a usual modules,
>>>but modules in a module.
>>>We need to address technical issues also. Technically,
>>>it is possible to separate session save handlers from
>>>session module.
>>>However, it does not mean, it's good idea to do that.
>>>
>>>If we decide to separate session save handlers to PECL,
>>>there must be *clean* way to separate them.
>>>
>>>If PHP has *clean* way to provide modules in a module,
>>>it will be realistic to place session save handlers
>>>to PECL. I expect changes in current *module*
>>>initilization code for this.
>>>
>>>Anyone has comment on this?
>>>(Compilation and initialization for modules in a module.
>>>We can put php_session.h in standard header location.
>>>Do we really want this? "Undefined Symbol" error is not
>>>acceptable at least for me, also. It's possible to have
>>>required globals always in PHP. Do we really want this?)
>>>
>>>BTW, currently, mm save handler support is broken partially.
>>>Clue for proper session save handler module
>>>initialization with current code is also appreciated :)
>>>
>>>
>>
>>
>>
> 



-- 
Yasuo Ohgaki


-- 
PHP Development Mailing List <http://www.php.net/>
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
To contact the list administrators, e-mail: [EMAIL PROTECTED]

Reply via email to