This is a rather inexplicable error message that Crystal Reports pops up. If you search for the message, you see a few ideas for solutions, but what I eventually did didn't seem to be on anyone's list, so I'm adding this post to the search in case it's helpful.
In my case, the big clue was that we had just modified the report viewer so it could pick up data from alternate data sources based on information in the web.config. It's a pull report, so the data source is specified directly in the report and we're modifying it programatically. My assumption was that somewhere in the report, there was something being specified to use a different data source than all the rest of the report, and that our code that changed the source was missing it. But I couldn't find it, and Crystal report files have a binary format, so they're next to impossible to search through.
The big clue was that we had just modified the report viewer so it could pick up data from alternate data sources
I dug through a book on Crystal last night - luckily it had a chapter on dynamic data sources; thank you Brian Bischof - and one of the things it suggested was to check the connections on the tables using a member function, TestConnectivity. (You can outfit each individual table that a Crystal Report draws upon with its own login information.) To create a dynamic data source, you have to make sure that the source is changed all over the place in the report. In the report object, in any subreports that it uses, in any tables. So I set up the table login information to set it the way it was supposed to be, and then call TestConnectivity, and if that failed, throw an exception. My idea was that there was a single table somewhere causing all the problems, and that if I could figure out which one it was, I could fix it.
Only I didn't need to.
As soon as I added the TestConnectivity check, the report started coming up.
My working assumption is that the changes to the table login information are cached somewhere, and don't actually take hold until the report thinks it's necessary. Presumably there's a bug in the caching system, but calling TestConnectivity causes the table login information to be set properly.
Nice weird one there. Spent too many hours on it though.