Re: [RFT] new store file/spool approach (fixed)

Kalle Marjola wrote:

> 
>> >  a) seems like the directory must exist already or bad things happen
>> 
>> hmm, store_init returns -1 if directory doesn't exists and it's up to
>> higher layer to check this (bearerbox init code). I don't like the idea
>> to create directory if it doesn't exists and also the idea to panic in
>> this case. IMHO we have to much panics spread across kannel for such low
>> level things and higher layer should check return values and make
>> decisions.
>> 
> However there is one difference: in old store, if the file did not exist
> that only meant that well there is no old messages; new ones were still
> stored. In this case nothing will get stored, but unless the user is
> observant, he won't notice this (or I'm missing something here)

hmm, I don't see difference... If user has enabled store-dir then 'static
Octstr *spool' is not NULL and we store messages... if while starting
spooldir doesn't exists we return -1 to higher layer and this should decide
how to process further. (or I'm on the wrong trip?;))

> (okay, I'm not sure what happens if the store file cannot be later on
> created.. just errors I guess..)
> 
> Hence, either panic or create the directory. But I can live with current
> behaviour, too..
> 
> 
> 
>> yep, good point... 'store-dir' is a good one?
>> 
> Yes, see below.
> 
> 
>> real benefit is that store spool dir is slower as e.g. store file + your
>> patch (dict approach) (if your gateway has enough RAM to store messages
>> twice). so user will make decision which type to use (but default:
>> store-dir :)).
>> 
> Well in that case I would suggest that we have both :]
> Which one is used depends on config setting:
> store-file : the old (and I will patch it with that previous patch)
> store-dir: the new one
> 
> ..and yell if both set...

yep should work too... would you please patch current/old bb_store and I
will create patch against this version?

> 
> 

-- 
Thanks,
Alex