0.15 to 0.16 checkpoint file might cause incorrect JSON output

[JamesHeinrich]

Unconfirmed report from a user: upgrading from mfakto v0.15(?) to v0.16-beta.2 with an in-progress checkpoint file (for Factor=690000037,79,80) resulted in a spurious factors entry in the JSON of a NF result, moreover the "factors" are an invalid hex string:
{"exponent":690000037, "worktype":"TF", "status":"NF", "bitlo":79, "bithi":80, "rangecomplete":true, "factors":[BC56A3A4], "program":{"name":"mfakto", "version":"0.16-beta.2", "subversion":"cl_barrett15_82_gs_2"}, ... "os":{"os": "Darwin", "architecture": "ARM64"}
To be clear, the "NF" result is correct, there are no factors to be found, but something in reading the old checkpoint file appears to confuse the new code into writing false output.

Perhaps there should be a sanity check when reading a checkpoint file to make sure it's in the expected format?
The report from the user:
I saw that the newer version of mfakto complained about the checkpoint files for that exponent and then mfakto ignored the checkpoint and started from the classes from the begining, so I thought mfakto handled it.

[ixfd64]

I'm not sure 0.16 officially supports older checkpoints. Unlike mfaktc, the params.h file for mfakto doesn't have a checkpoint file version identifier.

[JamesHeinrich]

Not supporting previous-version checkpoints is perfectly acceptable. What it should try to avoid, if at all possible, is reading in a checkpoint file and assuming it's current-version and stuffing invalid data into the factors key. If nothing else, when reading in a checkpoint file and appearing to come up with something to put in "factors", check to make sure that said factor(s) actually divide the exponent, and if not then reject the checkpoint file entirely as corrupt/invalid.
Although the user report seems to indicate that the checkpoint file was rejected, calculation started over, but the resulting JSON was still broken.

[ixfd64]

I checked the changelog and found this entry under version 0.10:

mfakto now allows reading checkpoints from any version of mfaktc and mfakto as long as the other parameters and the checksum are OK

So mfakto is supposed to be able to read checkpoints from previous versions as well as mfaktc. If it is unable to read the checkpoint for whatever reason, then I do agree there should be feedback to the user.