On Wed, 20 Dec 2006 00:19:26 +0100, Jan Kratochvil wrote:
> On Mon, 18 Dec 2006 21:09:19 +0100, Chet Ramey wrote:
> ...
> > Not exactly.  The callback and traditional calling mechanisms are
> > orthogonal.  It's just not safe to mix them.  The longjmp leaves
> > readline in an inconsistent state, unless the calling application is
> > careful to undo what state-setting it has done.  That's the part
> > readline can't know about:  the appropriate application-specific part.
> 
> OK, so going to provide callback based GDB reimplementation of
> gdb_readline_wrapper () (readline () call).

Here is one which forks for me but still it is more complicated than I expected.

I believe there may be side effects but unfortunately it can be expected the
interactive readline(3) usage is not well covered by the testsuite.
readline(3) call does a lot of various initializations currently omitted by the
callback-based gdb reimplementation.


Regards,
Jan