Re: 21.0-b47/pre4: bug in configure still exists.

Martin> I don't think this is right, since there are really only two database Martin> types, `dbm', and `berkdb'. gdbm is just another implementation of Martin> `dbm'. I think you can't have both `dbm' and `gdbm' actually in use Martin> within the same xemacs. This is even more confusing since I believe Martin> berkdb can also provide the `dbm' api. (not claiming to be a database Martin> expert, though). Gregory> I agree that at the moment, only two selections will work but the error Gregory> given would be incorrect for the command: Gregory> * ) USAGE_ERROR(["The \`--$optname' option value Gregory> must be either \`no' or a comma-separated list Gregory> of one or more of \`berkdb', \`dbm', or \`gnudbm'."]) ;; Gregory> It doesn't say one or two, it says one or more. If three are given it Gregory> gives this error and leaves the user wondering what they did wrong. After Gregory> the patch is applied, if they user types all three, they at least get an Gregory> error which makes sense: Gregory> USAGE_ERROR("Only one of \`dbm' and \`gnudbm' may be specified Gregory> with the \`--$optname' option.") Martin> Just don't expect to be able to get dbm and gdbm after you're done. Gregory> Never did, I saw the error following that case when I first ran across Gregory> this. OK, we are in complete agreement then. Thank you. Martin