At Sunday 08 August 2010, Ralf Wildenhues wrote: > * Stefano Lattarini wrote on Sun, Aug 08, 2010 at 07:42:39PM CEST: > > > The rest seems good for maint, thanks. > > > > BTW, the patch was against master, and while rebasing it I saw > > (well, a merge conflict made me see) that I could backport some > > extra checks in condinc2.test from master to maint. I did that, > > too. > > > > The updated patch is attached. > > You don't need another approval from me when I write that the patch > is ok with fixes. True, but in case I did more changes than just fixings (i.e. I backported some checks from master), and you disapproved this additonal changes. So asking for ACK of the updated patch was a good idea after all. > > Generally, however, I'd prefer to not get into the business of > backporting fixes from master by redoing them. OK, I'll keep this in mind: no test backporting unless there's a good reason.
> It has a technical > and a semantical downside: it makes git merge errors likely > (things end up doubled in the file), and one should introduces > changes once only; if they were added to master only, then either > there was a reason for that, or it was an error. Backporting > testsuite coverage by definition seems like a waste of time to me: > master is the interesting place for development, something that is > stable does not need more testsuite exposure. (This argument > doesn't hold for new patches against maint, as long as the extra > cost of applying it to maint rather than to master is almost zero; > if it isn't any more, then let's add the patches to master only.) Agreed. Regards, Stefano
