Re: [RFC] Fixing gdb.base/completion.exp (PR testsuite/12649)

[Jan Kratochvil <jan dot kratochvil at redhat dot com>]

On Mon, 02 May 2011 18:09:50 +0200, Pedro Alves wrote:
> This is about completion, using one form or the other.
> We could move the "\t" form to readline-completion.exp,
> but I think a systematic approach to testing all the
> completion methods is better, and helps maintenance in the
> long run.

I was addressing this by the readline/ and gdb/ parts testing differentiation,
the first paragraph of my mail.


> > I understand one cannot change the whole codebase to a better / more
> > maintainable form over night but when there are attempts and patches offered
> > IMO the current codebase should not be actively kept worse.
> 
> I took the time investigate the original issues with the code, write a patch
> to fix them, explain the problems and the proposed fixes, in order to not keep the
> knowledge to myself, and I've posted the beginnings of a patch that cleans
> up the test further.  I don't think it's fair to suggest I'm trying to keep
> anything worse.

There was the proposal using gdb_test "complete ...":
	Re: [RFC] Fixing gdb.base/completion.exp (PR testsuite/12649)
	http://sourceware.org/ml/gdb-patches/2011-04/msg00538.html
(I do not have that mail reviewed but it gives the idea.)

I consider the gdb_test "complete ..." testfile better than the (even fixed)
send_gdb-gdb_expect testfile.  Therefore I consider proposing the
send_gdb-gdb_expect testfile over the gdb_test "complete ..." testfile as
"actively trying to keep the testfile worse".

But there is the point that you do not consider the gdb_test "complete ..."
method having the same testing coverage as the "\t" testing method.  This is
the point where we do not agree and I agree in such case the gdb_test
"complete ..." change is not acceptable (for you).


Thanks,
Jan