Re: cov.m patch

[John W. Eaton]

On 12-Mar-2012, Jordi Gutiérrez Hermoso wrote:

| >> I'm wondering what to do about Ismael's cov.m patch... Ismael patched
| >> relative to 3.4.3... but I'm not sure if his patch is "serious" enough
| >> for stable. I merged it from 3.4.3 to 3.6.2, but I'm not sure if this
| >> really is a patch that should go on stable or if I should rebase it to
| >> default. I could also merge directly from 3.4.3 to default, but this
| >> leaves two heads on stable, and I think this might confuse other
| >> people.
| >>
| >> What I have right now is the patch on 3.4.3, merged into 3.6.2
| >> (current stable tip) and then merged again into default. AIUI, we're
| >> only patching regressions and serious bugs in stable. Is this bug
| >> serious enough? If not, I'll rebase, but I prefer to preserve the
| >> integrity of people's patches and actually put people's names next to
| >> the change they actually made, including the revision relative to
| >> which they patched.

I don't think this change is appropriate for stable.  It does not fix
a regression and it is a compatibility issue that we have discussed
previously and intentionally decided against following Matlab in this
case.  If we change that decision now, the place for the change would
be default, where we could get plenty of testing before the next
stable release.  I also think we should have a warning for this case
to let people know that the change in behavior is intentional.

What do you mean by patch on 3.4.3?  Do you mean you patched relative
to some old version of stable?  If the patch doesn't belong on current
stable, then I don't think it belongs on some earlier version either.

jwe