Re: [PATCH 0/11] Cleanup MI test cases

[Pedro Alves <palves at redhat dot com>]

On 09/28/2012 01:04 AM, Yao Qi wrote:
> Hi,
> This is a follow up to this thread,
> 
>   Re: [RFC] Merge mi-cli.exp and mi2-cli.exp
>   http://sourceware.org/ml/gdb-patches/2012-09/msg00155.html
> 
> in which we discussed that we are only using mi2 so far and some test
> cases are redundant.  This patch is to get rid of the redundancy in
> test cases.
> 
> Patch 01/11 simply removes mi-FOO.exp, because they are identical to
> mi2-FOO.exp.  The rest of the patches merge mi-foo.exp with mi2-foo.exp,
> and remove mi-foo.exp.

This is great, but I think it went the wrong direction.  :-/

We have more mi-* files and more changes to mi-* files than the
other way around:

ls mi-* | wc -l
54

grep-changelog --text "gdb.mi/mi-" ../ChangeLog | grep "^2" | wc -l
412

vs

ls mi2-* | wc -l
26

grep-changelog --text "gdb.mi/mi2-" ../ChangeLog | grep "^2" | wc -l
210

And that makes sense because the mi2-* files were originally created as
copies of the existing mi-* files, and the mi-* files supposedly continued
to evolve, until some point in the future when mi3 would be finalized (at which
point we'd again fork the mi3-* copies).  The mi2-* file were supposedly
cast in stone once the copying was done (on 2003-08-07).

Is it hard to redo things a bit to end up with only the mi- files instead?

I'd think that something around "mv mi2-foo.exp mi-foo.exp", plus
"git add mi-foo.exp", and maybe squash that commit back would get
you that effect without needing to really redo things.

> Note that this patch set doesn't cover mi2-var-child.exp and
> mi-var-child.exp, because they diverge too much for me to handle.

-- 
Pedro Alves