|
Hello newsgroup,
I am not sure, but I think I found a bug in libdvb
(0.2.2).
Frank Verlinden reported the following problem
in a posting dated 3 Mar 2003
> After loading the Convergence or the
Metzler DVB drivers the n-tv screen
> is showed. When starting Tuxzap and trying to switch to an FTA channel > the screen becomes blank. With the Convergence driver, it is still > possible to switch with szap. I had the same problem, but unfortunalety Frank did
not continue the thread,
so I tried to find out the reason and a solution by
myself. Let me now report
my results:
In source DVB.cc is a function set_diseqc_nb. Its
purpose is to
assign proper values to the six bytes and the
length attribute of
the struct dcmd, used as parameter for the frontend
API function
FE_DISEQC_SEND_MASTER_CMD later on.
Here is a part of the C code:
------------------------------
dcmd.msg[3]=0xF0 |
((nr * 4) & 0x0F) | (tone ? 1 : 0) | (voltage ? 2 : 0); ------------------------------------------------------------------------
According to the DiSEqC specification bus_spec.pdf (found at
http://www.eutelsat.com/satellites/4_4_5.html),
chapter 8.3 "Command Byte" and chapter 9.1 "Write Port Group", and
slave_microcontroller_spec.pdf, chapter 7.2 "Committed
Control Pins",
dcmd.msg[3] is the data byte in the
"Write N0" command (hex code 0x38, "write to port
group 0
(committed switches)").
The meaning of the least significant bit in this
data byte is:
1 <=> "High L.O. frequency
selected".
Now lets look at what happens if tuxzap switches to
a transponder
in the high band area:
In function DVB::SetTP tone is properly set to
SEC_TONE_ON,
but unfortunately
the enum type of tone is:
------ linux/dvb/frontend.h
---------------------------------
typedef enum fe_sec_tone_mode
{
SEC_TONE_ON, SEC_TONE_OFF } fe_sec_tone_mode_t; --------------------------------------------------------------------- (see also "Linux DB API Version 3", V1.0.0
pre1,
chapter 2.1.7 "SEC continuous tone")
That means, that "tone" has _zero_ value and
"(tone : 1 : 0)"
evaluates to 0,
resulting in the least significant bit in
dcmd.msg[3] NOT set
in this case.
This is a contradiction to the eutelsat
specs.
I patched the expression to: ((tone == SEC_TONE_ON)
? 1 : 0),
and a very similar piece of code in function
set_diseq the same
way. This is
much more robust coding, and it worked properly.
=============================================
Is this a bug?
I am doubtfull, because only one posting
reported this problem.
I would highly appreciate a statement of
Marcus Metzler.
Thanks in advance for any answer.
Best regards,
Sven Severus
|