Hi Colin,

At 2025-11-15T14:51:30-0800, Collin Funk wrote:
> [GBR wrote]:
> > What I took away from the 1.22.4 and 1.23.0 release cycles shows was
> > that 2 weeks is about as long as we need for RC feedback.  I
> > remember being frustrated that holding the RC window open for a
> > third week didn't seem to accomplish much.
> >
> > It's my first time doing this officially for groff, of course.  I'm
> > open to feedback or suggestions about release management.
> 
> For Coreutils, Pádraig usually does a release snapshot around a week
> before the planned release to allow people to test [1].

I noticed.  :)  I'm subscribed to the coreutils mailing list.

I think the differences in the nature and size of the audiences for
these two different projects motivates a longer evaluation timeline for
groff, judging by the chronology of feedback on RCs I've seen in the
past 2 groff releases (1.22.4, 1.23.0) and the last coreutils one (9.9).

One thing the projects have in common is that they offer a variegated
basket of command-line programs, of which some get used frequently and
directly at the command line (in groff's case, _almost_ directly, via
man(1)) and others of which are relatively esoteric, with few users
outside of scripts.  I imagine the latter to characterize numfmt(1) and
afmtodit(1), for example.

> Sometimes it goes well and can be released early. Other times it needs
> a few more snapshots due to unforeseen bugs.

The way I see it is: you gather reports of defects in the RC, categorize
each (valid) issue either as requiring immediate remedy or as deferrable
to the next release.  If the count of "immediate remedy" defects is
greater than zero, you roll another RC.  If the count of deferrable
defects is nonzero, you describe them in the release notes.  When the
count of defects demanding immediate remedy is zero, release!

> Just figured I would share in case you want to try something similar,
> using [email protected].

We availed ourselves of "platform-testers" in the groff 1.23.0 release
cycle and its members were indeed helpful.  I do plan to do so again.

https://cgit.git.savannah.gnu.org/cgit/groff.git/tree/ANNOUNCE

For 1.23.0, I functioned as "deputy maintainer/release manager" because
our maintainer, Bertrand Garrigues, had sharply limited availability.
This'll be my first time with a full bird on my lapel.

Regards,
Branden

Attachment: signature.asc
Description: PGP signature

Reply via email to