Serials predictions fail when nominally omittable chronology caption fields not in pattern

[Lebbeous Fogle-Weekley]

In Evergreen master (and probably down through 2.0):

Sorry about the title of this bug. There is simply no way to talk about how serials predictions work in plain English.

Olli-Antti Kivilahti reported trouble creating serials predictions in the #evergreen IRC channel.

He was trying to set up a weekly with any of the patterns in the following paste, without success: http://pastebin.com/CN8vFiyP

I have suggested the alternative in the following paste, which does work, but it may not be exactly what the user intended: http://pastie.org/3727277

The relevant difference is the presence of ‡i and ‡j in my pattern. You *should*, however, be able to omit those. But if you do, the middle layer routines for prediction seem to hang forever, and don't seem to emit error messages. Investigation is called for. I'll try to do what I can, but help from all sources is appreciated.

[Dan Wells (dbw2)]

I have made some progress on this and am currently testing my fix. Barring unforeseen issues, I should have a branch for this tomorrow.

[Dan Wells (dbw2)]

Here is the branch with the fix (just one commit):

http://git.evergreen-ils.org/?p=working/Evergreen.git;a=shortlog;h=refs/heads/user/dbwells/lp973453_incomplete_chron_bug

I think this can (and should) backport cleanly through 2.0. If it doesn't, just let me know and I will take care of it. Thanks!

[Lebbeous Fogle-Weekley (lebbeous)]

Thanks Dan! I aim to test this by the end of today.

[Lebbeous Fogle-Weekley (lebbeous)]

Thanks again Dan. Sorry it took longer than I said, but this is now fixed and backported through rel_2_1. Looking into whether rel_2_0 will work now.

[Lebbeous Fogle-Weekley (lebbeous)]

rel_2_0 handled as well

All done!