Sample-session: (was playing around while reading the docs)
$ tclkit-linux-x86
% package require Mk4tcl [-> 2.4.9.1 ]
% mk::row create foo bar bar snafu [-> _._!1._!0 ]
[ditto for a few repetitions, then...]
% mk::row create foo bar bar snafu [-> SEGFAULT]
$ echo $? -> 139
The number of repetitions till crash varies. (the machine
itself usually runs very reliably)
The returned row-id is always the same (correctly so,
since no cursor is bound to it)
-- [email protected]

I suspect that this is yet another bug related to MK's attached vs. unattached
views. The
best advice I can give is to stay away from standalone rows as much as possible
(I
usually keep intermediate data as Tcl lists).
It won't happen when you do "mk::append db.myview ...", which is attached, in the
sense
that the data goes into a view which is part of a datafile.
This is nto likely to be resolvable soon - it goes to deep into MK internals,
and the
Mk4tcl design makes it even harder to get it right. Solving the core issue (attached
vs.
unattached views) is a top issue for the next MK design, but this will clearly
not be of
any help to you right now.