RFA: New flag in struct rtx_def

[Richard Earnshaw]

> >>>>> "Richard" == Richard Earnshaw <rearnsha@arm.com> writes:
> 
>     >> That seems unfair; it's taken a while because we've been
>     >> talking about how properly to fix the problem.
> 
>     Richard> Not entirely unfair.
> 
> I understand your frustruation, but I think that probably Jason didn't
> realize how severely inconvenienced you were.  Jason is now working on
> solving the problem; I'm sure he will do so shortly.
> 
> Let's not criticize each other personally.  (Technical criticism is
> fine, if polite.  For example, "I think a better solution to problem X
> would have been Y".)  If anyone feels that a contributor or maintainer
> isn't acting appropriately, let's either handle that with private
> mail, with mail to the SC, or with mail directly to me, if preferable.
> 
> I see it as part of my job to keep things building, etc., and I'm
> happy to handle these kinds of management issues.
> 

My apologies, I should have chosen a different way of expressing that.

The frustration lies in the fact that this isn't the first time that the 
ARM port has been left unbuildable for several days because of a 
front/mid-end change.  Just because there aren't fifty or whatever 
engineers banging away on that particular port of the compiler every day 
doesn't mean that it is any less inconvenient if it gets broken.

I don't expect that every checkin should be tested on every platform; that 
would be completely impractical.  But I do think there should be a 
commitment made when a change is made to a MI part of the compiler that if 
it does cause problems for a particular port it should be looked at as a 
priority and, if a resolution is not readily available, consideration 
should be given to reverting the patch until such time as a solution can 
be found.  Clearly this is a two way issue, since it will often require 
the assistance of back-end maintainers to understand why a particular 
change is problematical, but equally, individual maintainers won't 
necessarily understand why a front end change was made.

Richard.