Incorrect encoding of version M1/M3 tail data code word

[vlad417]

The M1 and M3 versions' encoding of the 4-bit last code word of the data appears to be incorrect. It looks like it is always shifted right by one (e.g. a code of 14 is encoded to read 7). It appears that qrencode.c:ln587 (bit = 1 << raw->oddbits;) in QRcode_encodeMaskMQR should read bit = 1 << (raw->oddbits-1);

It functions correctly with that change as far as I can tell. I apologize for not simply committing the change, but I do not have git on my work PC.

[fukuchi]

Thank you very much vlad417, the bug is confirmed. You must have hawk eyes! If you have a test pattern to identify this bug, would you mind to send me it? I'll add it to the test code. I just need an input string and an expected symbol.

[vlad417]

No problem. I noticed the discrepancy against our decoder. Confirmed it by writing the frame out to an image right after the "inteleaved data and ecc codes" block and decoding that by hand. Happened with the following set-ups:

ver M1, byte values: 52 54 49 57 52
ver M3-M, byte values: 87 66 65 53 89 52 55 89 80 81 81

[fukuchi]

Done. Thank you for your contribution.