Date problem corrupted record

[Wes Groleau]

Trying to test enhancement to date picker with SVN 9782.
Created new INDI and pasted in events with various date formats.
Trying to save the changed caused a stack dump in the edit_interface.
Used "Back" to recover the GEDCOM. Inspected it--all valid.
Took out all the events and tried to save--still a stack dump
Tried to delete the record--another stack dump
Refresh page--pending changes note appears (I have my changes set to automatic).
Went there, and see several Replace, then Delete.
Tried to accept all--stack dump
Reject all worked.

Made another INDI, and pasted in a few events at a time. Dates in the future are NOT a problem.

I've narrowed it down to nine events--I will do more testing to narrow it down more, but I wanted to enter this much so that I don't forget the details overnight.
The problem is definitely associated with dates, as the stack dump indicates. I think all my formats are legal. The dump is:

ERROR 8: Trying to get property of non-object
0 Error occurred on line 848 of file functions_import.php in function update_dates
1 called from line 668 of file functions_import.php in function import_record
2 called from line 1285 of file functions_import.php in function update_record
3 called from line 1192 of file functions_import.php in function accept_all_changes
4 called from line 401 of file functions_edit.php in function delete_gedrec
5 called from line 2594 of file functions_edit.php in function delete_person
6 called from line 1717 of file edit_interface.php

Notice: Trying to get property of non-object in /home/content/G/T/S/GTSCFL/html/UG/webtrees/includes/functions/functions_import.php on line 848

Fatal error: Call to a member function Format() on a non-object in /home/content/G/T/S/GTSCFL/html/UG/webtrees/includes/functions/functions_import.php on line 848

[Wes Groleau (webmaster-unigen)]

The following dates did NOT cause the problem, and they all looked OK in the display except as NOTEd.
Repeat, these have been ruled out--still looking for the bad one(s).
1 BIRT
2 DATE 9 AUG 2019
2 NOTE testing
1 CHR
2 DATE AUG 2019
2 NOTE testing
1 BARM
2 DATE 2019
2 NOTE testing
1 GRAD
2 DATE ABT 9 AUG 2019
2 NOTE testing
1 CENS
2 DATE INT AUG 2019
2 NOTE *** This one was displayed with a set of
3 CONT *** empty parens after the year.
1 EMIG
2 DATE BEF 2019
2 NOTE testing
1 IMMI
2 DATE @#DGREGORIAN@ 9 MAY 101 B.C.
2 NOTE testing
1 NATU
2 DATE @#DGREGORIAN@ 9 MAY 101
2 NOTE testing
1 FCOM
2 DATE @#DGREGORIAN@ 9 MAY 1001 B.C.
2 NOTE testing
1 BLES
2 DATE @#DGREGORIAN@ 9 AUG 1698/99
2 NOTE testing
1 ORDN
2 DATE 09 AUG 1698/99
2 NOTE testing
1 OCCU
2 DATE @#DGREGORIAN@ 9 AUG 2019
2 NOTE testing
1 PROP
2 DATE EST @#DGREGORIAN@ AUG 2019
2 NOTE testing
1 RELI
2 DATE AFT @#DGREGORIAN@ 2019
2 NOTE No record corruption so far

[Wes Groleau (webmaster-unigen)]

It is "UNKNOWN"
Granted, it would be pretty unusual for someone to enter that, but still, webtrees shouldn't crash. I could see someone using it for one the spec doesn't provide for.

The following event, added with nothing else to a just-created INDI, causes the problem:
1 BURI
2 DATE @#DUNKNOWN@ 9 AUG 2019
2 NOTE SVN 9782 chokes on this one
----------------------------------
For what it's worth, the rest of these are verifed SAFE:
1 RESI
2 DATE BET 9 AUG 2019 AND 01 APR 2038
2 NOTE testing
1 TITL
2 DATE FROM 9 AUG 2019 TO 9 AUG 2119
2 NOTE testing
1 NATI
2 DATE BET @#DGREGORIAN@ 9 AUG 2019 AND 01 APR 2038
2 NOTE testing
1 EDUC
2 DATE FROM @#DGREGORIAN@ 9 AUG 2019 TO 9 AUG 2119
2 NOTE testing
1 ADOP
2 DATE @#DJULIAN@ 9 AUG 100 B.C.
2 NOTE testing
1 RETI
2 DATE BEF @#DHEBREW@ KSL 2019
2 NOTE testing
1 DEAT
2 DATE @#DFRENCH R@ 9 BRUM 7
2 NOTE testing
1 PROB
2 DATE INT @#DROMAN@ 9 AUG 2019

[Wes Groleau (webmaster-unigen)]

P.S. If the INDI existed before the crash, "reject changes" instead of rolling back to that state, deletes the INDI and then creating a new INDI does not re-use the number.

[fisharebest (fisharebest)]

Fixed in SVN 9808

The invalid calendar escape was causing the constructor to drop out of a switch() statement.

As a result, the date object was created, but not initialised. This causes problems with any code that tries to use this date object.

[fisharebest (fisharebest)]

@#DGREGORIAN@ 9 AUG 1698/99

PS: although the gedcom specification describes old-style/new-style years as "gregorian", they are only used on julian dates.

And, of course, the OS/NS notation is only valid for dates between 01 JAN and 24 MAR ...

[Wes Groleau (webmaster-unigen)]

"UNKNOWN" _is_ one of the options in the 5.5.1 spec.
I doubt anyone will ever use it, but I was trying to make sure my date-picker effort would not crash on anything the spec allows.

[fisharebest (fisharebest)]

Fix released in webtrees-1.0.6