Was this file converted from a pre-7 version?
I have encountered something similar with blank but ineditable and undeletable Value List definitions in a file which was started around FM3 and converted to FM7 years ago.
The only answer anyone ventures was "file corruption" which is a nice catch-all which doesn't help much. Especially with a huge multifile system that was converted years ago and has been running without issue on our Server inspite of this previously unnoticed issue in one file out of more than 60.
I suggest taking a copy of the problem file off-line (onto another computer) and run the recover there, not with the idea of replacing the file or using the recovered file, but to see what the recovery log can tell you.
(Even if it says nothing is wrong, trash the recovered file after opening it to see what happened to the data reference. I refuse to believe that recovered files are ever reliable; the recovery process isn't intended to make the reliable, just to make the data recoverable, thus the process name Recover, not Fix.)
Let us know what if anything you learn in the process.
I have the same thing in one of my databases. It was originally built in FMPA10 and have not been able to edit or remove it.
File is hosted on FMSA11, it is the only database on that server with the issue.
Doesnt appear to be causing any harm so i've just left it there.
My file was created in FMA-11.
Unfortunatelty, I do not have a clone from a time when this symptom did not exist. My last copy of the database that does not include this problem is December, 2011.
Yes, running Recover does remove the unwanted file reference, but the Log does not tell me where it comes from or how to remove it. I know better than to use the Recovered edition in production use.
I was hoping there might be another way to remove it or edited, perhaps by tricking the graph into using it or something like that.
1. Can you compare the Dec 2011 backup list of Data Refs to the problem file and get any clue which one of the December Refs has "gone bad" since then?
I had not heard of this type of thing in files built with newer versions. I am wondering if the removal of obsolete file TOs and the actual removal of the source file from a server could have cause something like this when the Data Ref target no longer exists.
I mention this because we did a lot of file clean-outs from our server after upgrading to 11.
2. Any idea of such deletions from the wider system were done that might have included anything in the old file's Data Refs?
Meanwhile, while I don't like it, I've done what Matt did with my own problem file -- left it in service as it is working fine (for now).
I went ahead and did the same as dburnham.
I took a clone of the file from Dec 2010 that still had the empty Data Source. Ran a recover and it stated no issues were found, but the empty data source is now gone. Nothing in the log indicated anything.
I ran a DDR on my current file for just data sources. It shows i have 3 ( 2 of which are those i created) and the third just says <Missing Data Source>
Is everything gone from that datasource: any layouts, any table occurrences on the graph, any references in other tables (via calc or lookup), in value lists? in custom functions?
Beverly: no, nothing at all is missing. The database is otherwise appearing to function normally.
Stephen: There is no clue from the December 2011 edition. At that time, there were two external references. Now, there are three, including this "phantom" one that refuses to be edited or deleted. I suppose at some time after December, something was added and perhaps removed, but my server backups only go back to the usual FM Server cycle.
Wow! Something created or left a bad data reference more recently that the FM 11 version change...
I hadn't even considered the possibility that my bad Value List or your phantom Data Ref were due to something so recent, unless somehow they got orphaned from a test TO which was deleted without it cleaning out dependant stuff properly.
Makes me wonder about sutff like having the TO graph open at the time a backup begins, and similar schema stuff that's nearly impossible to test. I try to be careful about editing schema around backup times, but I know it has caught me unaware a few times in the last couple of years.