hei Devin, Thank you for debugging this issue. Devin Heitmueller wrote:
Thanks to Jelle providing an environment to debug the issue in, I isolated the problem. This is actually a combination of bugs in mplayer and the af9013 driver not handling the condition as gracefully as some other demods. First the bugs in mplayer: The following is the line from the channels.conf where tuning failed: Frequency in question: 3FM(Digitenne):722000000:INVERSION_AUTO:BANDWIDTH_8_MHZ:FEC_AUTO:FEC_AUTO:QAM_AUTO:TRANSMISSION_MODE_AUTO:GUARD_INTERVAL_AUTO:HIERARCHY_AUTO:0:7142:1114 Mplayer does not support "TRANSMISSION_MODE_AUTO", "GUARD_INTERVAL_AUTO" and "QAM_AUTO" (for the constellation). In the case of the transmission mode and constellation, mplayer does not populate the field at all in the struct sent to the ioctl(), so you get whatever garbage is on the stack. For the guard interval field, it defaults to GUARD_INTERVAL_1_4 if it is an unrecognized value. I confirmed the mplayer behavior with the version Jelle has, as well as checking the source code in the svn trunk for the latest mplayer. So, why does it work with some tuners but not the af9013? Well, some demodulators check to see if *any* of the fields are "_AUTO" and if any of them are, then it puts the demod into auto mode and disregards whatever is in the other fields. However, the af9013 looks at each field, and if any of them are an unrecognized value, the code bails out in af9013_set_ofdm_params(). As a result, the tuning never actually happens. The behavior should be readily apparent if you were to put the above line into your channels.conf and try to tune (note I had to add printk() lines to af9013_set_ofdm_params() to see it bail out in the first switch statement. Anitti, do you want to take it from here, or would you prefer I rework the routine to put the device into auto mode if any of the fields are auto?
af9013 is correct in my mind. af9013 will return -EINVAL (error invalid value) in case of first garbage value met (maybe better to switch auto mode when garbage value meet and print debug log?).
Of course there should be at least debug printing to inform that... but fix you suggest is better for compatibility. You can do that, it is ok for me.
Devin
Antti -- http://palosaari.fi/ -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
