Re: MI testsuite failures [PATCH]

[Daniel Jacobowitz <drow at false dot org>]

On Wed, Jan 10, 2007 at 08:09:38PM +1300, Nick Roberts wrote:
> This doesn't seem a sensible way to do it as eventually we'll have massive
> duplication (multiplication?).  Note that the manual claims that GDB supports
> mi1 but there don't appear to be any tests for it.  I think:
> 
> 1) mi-*exp should be tests that work for all mi`N' where N > 1, and could be
>    run for each MI interpreter.
> 
> 2) mi2-*exp tests that work for mi2 but fail for mi3, currently none.
> 
> 3) mi3-*exp tests that work for mi3 but fail for mi2, currently none.

Running a single .exp file for multiple MI interpreters would be a bit
tricky, but we could probably do it.  You'd have to wrap most of them
in a function and call it twice, I guess.

I don't feel strongly about this if you want to change it.

> Also if mi3 becomes the new mi, we presumably should advise frontend
> developers to specify mi2, otherwise existing frontends might get a nasty
> surprise when the new features of mi3 appear.

I haven't thought about it much.  They'll appear at least one release
before we change -i=mi though.

> Hmm, what is this incompatible change anyway?

Quoting.  I posted an analysis of command line quoting issues to the
gdb list around the middle of last year; I intend to make every single
MI command handle quoting the way the manual says MI ought to, but this
will change the behavior of various commands (the directory and file
related ones, mainly, but -gdb-set will probably be affected too).

-- 
Daniel Jacobowitz
CodeSourcery