Re: mark_object crash in 22.1 and latest CVS (as of tonight)

[Richard Stallman]

    If I understand correctly, this means that either markers used for
    saving match-data should not go to `buffer-undo-list' or the "evaporate"
    option set by `save-match-data' is inherently broken.

I think `evaporate' is broken.

We could arrange a way to avoid putting these markers in
`buffer-undo-list', but then they would not be relocated correctly if
the body does some undoing.  So that is not a solution.  And once the
marker is in `buffer-undo-list', the Lisp code can get ahold of it.

The simplest solution is to make `evaporate' not do any thing special.

Here's a second solution.  Define a flag bit in markers, and set that
bit when relocation of the marker puts it into `buffer-undo-list'.
`set-match-data' can free the marker if the bit is not set.