commit 496862818d2feabcac926a94a6be2d42826ab19f
author Neels Hofmeyr <neels@hofmeyr.de> Wed Dec 05 21:32:21 2018 +0100
committer Neels Hofmeyr <neels@hofmeyr.de> Mon Dec 10 13:18:38 2018 +0100
tree ab1ec9bbb019096500de5c1aedf5cc8cf300e986
parent 2ca8cebac67cfa179af77aa8d507fd4b96b2b230

gsm0408_test: test encoding and decoding Mobile Identity

One would think by now we would solidly encode and decode Mobile Identities.
Well, guess again.

- rc is sometimes the amount of bytes written, sometimes actual strlen().

- on string truncation, rc is sometimes strlen() (assuming nul terminated), and
  sometimes snprintf()-style would-be strlen().

- returned string, when truncated by not enough buffer size, is sometimes nul
  terminated, sometimes not.

- gsm48_mi_to_string() happily reads a byte from zero-length input buffer.

- gsm48_mi_to_string() happily writes to zero length output buffer.

- gsm48_mi_to_string() returns nonempty string for empty input.

- encoding a MI type that still has the GSM_MI_ODD flag set results in encoding
  an even-length MI as odd-length (hence appending a stray 'F').

I am going to tweak the implementation of gsm48 mobile identity encoding /
decoding, so first pinpoint the current behavior in a unit test, and show how
perforated even such a seemingly trivial API can be.

Change-Id: Iaae3af87f82f1a8f2e6273984c011b2813038cf7