Modem SORACOM (SC-QGLC4-C1) : Unable to find a usable modem character set

[Frederic Martinsons]

Thanks for the answer Aleksander, by the way, I finally got a nano sim for my model and all went well. But, sadly same errors happens if the sim is locked (same error on CSCS than when the sim was missing).

I'll try at the beginning of next week to implement what you suggest on default encoding (by the way, what would be the best default: GSM, UTF8 ?)

Do you prefer that I issue one MR for mmcli protection about peek_modem and one for default charset? Or can I do both on the same MR?

Have a good week end!
________________________________
From: Aleksander Morgado <aleksander at aleksander.es>
Sent: Saturday, December 19, 2020 10:40:32 AM
To: Frederic Martinsons <frederic.martinsons at sigfox.com>
Cc: modemmanager-devel at lists.freedesktop.org <modemmanager-devel at lists.freedesktop.org>
Subject: Re: Modem SORACOM (SC-QGLC4-C1) : Unable to find a usable modem character set

Hey!

>
> After a sim was inserted in the dongle, no problems have been notices so I it seems that a missing sim prevent modem initialization (which, to my opinion should not be , it should failed only on enable step).
>

Yes, I agree. If the CSCS=X operation fails, we should still go on,
maybe assuming a default charset encoding. For this specific setup,
the modem will end up in failed state anyway, so it would be fine.

> Nevertheless, I open a bug (https://gitlab.freedesktop.org/mobile-broadband/ModemManager/-/issues/289) to track the mmcli segfault on a badly initialized modem and I'll try to come up with a MR to fix this (I don't know yet if the real fix should be to not expose the modem on DBus when it is not initialized or simply add a null-protection somewhere in mmcli)

The real fix is in ModemManager daemon; it looks like the daemon is
publishing a MM object in DBus, BUT which does NOT implement the
"Modem" interface, and that is completely wrong. The MM objects in
DBus must always implement the "Modem" interface. That said, yes,
mmcli can also be improved so that we don't just assume the "Modem"
interface always exists in the MM objects. If you're up to fixing
this, the problem is that mmcli is assuming the return of
mm_object_peek_modem() or mm_object_get_modem() is always non-NULL.

--
Aleksander
https://aleksander.es
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/modemmanager-devel/attachments/20201219/a91465fc/attachment.htm>