|
The current version
of the smsc_at code uses a rather brute force approach to opening and using the
serial port in that it basically tries for a finite amount of times to open the
interface, and then if this fails it does a number of retries. What this means
is that if the modem crashes, the serial port gets unplugged or various other
scenarios happen then the whole thing locks up and we hit
problems.
The modifications
we made to it recently (yet to be merged into CVS) apply a little bit more
intelligence by having various levels of drop back on failure, however these are
still based around the 'it will work at somepoint' mind-set rather than the
'expect the worst' (see earlier posts from me to get a copy of the patches if
you want a look).
Ideally the whole
of the smsc_at module could do with a good overhall to make it asyncronous and
to learn from the behaviour patterns that everyone in this group has worked out
(as well as to make it upgraded to the new smsc code style). Best plan
would be to put a team together to work on this part of the code and it's
architecture - I suggest Nick Clarey, myself, yourself, any other volunteers -
and then we can ideally produce a fully working at smsc gateway (*smslink and
all the others in the market that I have found have various other limitations so
it would do everyone a favour to do this).
Let me know what
everyone thinks.
Alex
skywire.co.uk
|
- smsc_at question Alexei Pashkovsky
- RE: smsc_at question Alex Judd
- RE: smsc_at question Kalle Marjola
