This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: PR 9479 (was Re: GCC-3.2.2 pre-release (third iteration))
On Friday 07 February 2003 14:36, Daniel Jacobowitz wrote:
> On Fri, Feb 07, 2003 at 11:15:52AM +0100, Franz Sirl wrote:
> > That's exactly the reason why I regularly request the binutils mainline
> > version to be bumped automatically when a release out of the branch is
> > done. Then it's much easier to correlate the binutils versions. Strangely
> > enough nobody else seems to feel the need...
>
> Because:
> - I keep suggesting you ask on the binutils list, rather than nagging
> people in private.
Uhm, I asked on the list without much response, so I kept quiet since then. I
picked up the ball again only because this situation fits :-).
> - In coordination with releases isn't the right time to do it.
> Whenever there's a relevant feature, that's the right time to do it.
> The releases come from a different branch, and what works in one may
> not in the other.
The point I'm trying to make is about "has at least these features working", I
don't care if the current binutils mainline has more working features or
fixed bugs.
So for example to check for 2.13.1 or better features with the mainline bumped
to 2.13.91 the check would look like:
(relver >= 2.13.1 || mainlinever >= 2.13.91 || hjlver >= 2.13.91.0.0)
Now this translates into quite simpler pattern matching (mainlinever and
hjlver can easily be combined into one match) compared to now. And every time
when a feature/bug fix requiring a date match is backported to the branch and
subsequently released, the date checks can be replaced with simple version
checks.
So, why not bump the mainline version on every feature addition/bug fix?
Simply practical reasoning, as with most things that everybody is responsible
for, in the end nobody will actually do it. So this would require some kind
of "version bump manager", which sounds like overkill to me. Also this will
actually give us more granularity than we usually need, As long as binutils
releases happen in regular intervals (Thanks Daniel!), bumping mainline
version on every release sounds like a simple and realizable plan to me.
As for the feature checks that Joe suggested, history shows that a lot of
binutils feature checks are next to impossible to implement in a
cross-platform way, that's the reason why the version and date checks are
there after all.
Franz.