This is the mail archive of the
fortran@gcc.gnu.org
mailing list for the GNU Fortran project.
Re: 4.2 Backport fix for PR libfortran/31196
On Mon, Apr 16, 2007 at 10:58:58PM +0000, Joseph S. Myers wrote:
> On Mon, 16 Apr 2007, Steve Kargl wrote:
>
> > 2) To address this problem, I briefly described how I think the
> > developement could be improved. 1) Mainline is branched for
> > release X.Y.0. 2) Mainline is closed to all commits while the
> > all effort is directed at getting X.Y.0 released. If everyone
>
> At this point, with regression fixes going on X.Y branch only and not on
> mainline, you end up with regressions in the next release because someone
> forgets to apply their patch to mainline when it reopens after X.Y.0 was
> released.
Since mainline and the X.Y branch can only differ by the patches
committed to X.Y, the release engineer can svnmerge those commits
from the X.Y branch into mainline prior to opening mainline for
development.
> The natural alternative to avoid such regressions would be to have
> mainline in release branch mode right up until the release, so release
> candidates come from mainline and branching is immediately followed by
> releasing with no intervening commits to the branch. (Or, indeed, to
> release from mainline and only branch some time after the release.)
> Given the length of time mainline spent in release branch mode for 4.2
> anyway, I don't think that would help get releases out.
This is the current timeline:
|
+-- GCC 4.2 branch created ------+
| (Oct 20 2006) \
v v
GCC 4.3 Stage 1 (ends Jan 20 2007) GCC 4.2.0 release
|
v
This is the proposed development model
|
+-- GCC 4.2 branch created -- (20 Oct 06) ----+
| |
Mainline Frozen (20 Oct 06) |
| v
| GCC 4.2.0 release (4 Nov 06)
Mainline Unfrozen (4 Nov 06)
|
v
GCC 4.3 Stage 1 (ends Jan 20 2007)
This model can work if all the cats can be herded for 1 or 2 weeks.
--
Steve