FC project direction

[Spacefalcon the Outlaw]

Das Signal <das.signal at freecalypso.org> wrote:

> I strongly agree that Pirelli and Compal are a dead end; targeting GTA02
> is a good idea considering it's much closer to the reference design and
> you have obtained the layout. Of course developing new hardware is the
> best course of action imho.

I'm glad we are in agreement; it just bites that with this "proper"
plan of development we are so far away from gratification in the form
of any actual usable phone...

> I don't currently own a GTA02 (yet),

If anyone here happens to have a spare Freerunner you aren't using,
would you consider donating or at least lending it to DS?  Right now
DS is the only member of our community besides me with some developer
qualifications, and if he had a device that could run both TCS211 and
our work-in-progress FC GSM fw, I feel that he might be able to help
us debug the breakage in our fw relative to working TCS211.  DS or any
other developer wishing to take on this task would need either a GTA02
or our own proposed dev board, but the latter is at least several
months away, realistically speaking, hence having someone donate a
spare GTA02 to DS would be the most expedient way to help the project.

(Yes, the lack of JTAG access on GTA02 is unfortunate, but if the fw
 hasn't crashed hard, there are ETM commands accessible via fc-tmsh to
 read and write any memory location - and the current fw problems are
 high level misbehavior, not hard crashes or hangs.)

> Unfortunately I have no knowledge about
> board design, except that directly connecting Vdd to Gnd is a bad idea ;)

Perhaps it would make sense then for me to switch my focus from the fw
to the board design subproject some time soon (maybe right after I
implement the AT-over-RVTMUX feature), and while I work on the board
design, give DS and others a chance to try their hand at debugging the
currently non-working fw.  If DS can obtain a Freerunner and move the
fw bring-up further along even a little bit while I work on the board
design, we will have achieved at least a little bit of parallelism. :)
And if neither DS or nor anyone has any success in advancing the state
of the fw while I work on the hw, oh well, the total time for me to do
all of the work should be the same no manner which order I do it in.

In fact, I just thought of another reason why it might be a good idea
for DS to get a GTA02 and try his hand at debugging the fw bring-up:
if I'm not mistaken, DS has a working OpenBTS setup, so he could see
what the network sees when the fw is failing to behave correctly.  I
don't have that setup, so I have to do all of my testing against a
live commercial network, which is not the most socially responsible
thing to do...

> To attract more developers, an idea could be to succeed at an initial PoC
> port to C1xx,

If this "proof of concept port" means having our fw run on C1xx (or
Pirelli, a slightly easier stepping stone) without any LCD or keypad
drivers (display stays dark, buttons do nothing) with control via AT
commands over RVTMUX presented on the sole easily accessible UART,
creating such a port would indeed be very little extra work beyond
what we have to do anyway for our main line of development, so it
would be imprudent not to do it.

But the problem is that I don't see a way to do it in a usable manner
*right now*, without getting our fw working like leo2moko on the FR
first.  Right now the main blocker that keeps our gsm-fw limited to
the gtamodem target only is the tpudrv12 module.  Suppose I were to
invest some time right now into extending that tpudrv12 code to
support Compal/Pirelli RFFE signal wiring in addition to the current
Leonardo/Openmoko one, and we got the fw running on Compal/Pirelli
targets in addition to gtamodem.  But what would we accomplish by
doing so?  Given that the fw currently *does not work* properly on its
native target, surely it will be just as broken (or likely more broken)
on these alien targets.  The only way I know of how we can debug it
and bring it into a working condition is by detailed action-by-action,
step-by-step comparison against working TCS211 - and trying to port
that blob-laden, Windows-built TCS211 to C1xx is NOT something I
relish doing.  Thus a Freerunner is still needed for debugging and
comparison, hence getting our gcc-built fw running on C1xx/Pirelli
ahead of making it work properly on gtamodem seems to accomplish
nothing.  But if you see some flaws in my reasoning here, please point
them out.

Hasta la Victoria, Siempre,
SF