Re: PATCH to use cpplib's source_location for location_t

[Mark Mitchell <mark at codesourcery dot com>]

Richard Henderson wrote:

> Ok.

I am not happy with the idea of checking in new features with known 
bugginess.  If anyone uses this stuff, they will get burned, and if Per 
doesn't finish it, we will drag it around for a while and throw it out.

I am a past offender here; the early RTL-based attempt at SSA, for 
example, sat around confusing people for a long time until it was 
finally killed.

The major arguments in favor of this approach seem to be that (a) having 
it in the tree gets people to use it and therefore provide feedback, and 
(b) makes it easier for the submitter by not having to do merges.

I consider (b) specious; there is a cost to everyone else from having 
the code in the tree, and it's not clear to me that minimizing the 
submitter's cost should be our objective.

I don't consider (a) specious, but the benefits needs to be weighed 
against the possibility that the work will not be completed and that 
until it is the presence of the code and associated options will be 
potentially confusing to developers and users.  It's not really that 
hard to check out a branch if you want to play with a feature.

--
Mark Mitchell
CodeSourcery, LLC
mark@codesourcery.com