On Tue, Jun 3, 2025 at 10:46 PM Richard Sandiford via Gcc <gcc@gcc.gnu.org> wrote: > > David Edelsohn via Gcc <gcc@gcc.gnu.org> writes: > > On Tue, Jun 3, 2025 at 3:23 PM Richard Sandiford <richard.sandif...@arm.com> > > wrote: > > > >> David Edelsohn <dje....@gmail.com> writes: > >> > On Tue, Jun 3, 2025 at 6:22 AM Richard Sandiford via Gcc < > >> gcc@gcc.gnu.org> > >> > wrote: > >> > > >> >> Hi, > >> >> > >> >> At the moment, all reviewers and maintainers have to be appointed by the > >> >> Steering Committee. I wonder if we could add a second, more > >> >> community-based > >> >> route: someone can be appointed as a reviewer or maintainer with the > >> >> agreement > >> >> of a given number of people who already have an equal or greater remit. > >> >> > >> >> It's already possible for reviewers or maintainers to defer to the > >> >> opinion of someone they trust and rubber-stamp that other person's > >> >> review or patch. Having the ability to appoint the other person as a > >> >> co-reviewer or co-maintainer of that area is really just replacing > >> >> patch-by-patch trust with a more ongoing trust. > >> >> > >> >> If that seems a bit woolly, and if a more formally defined process > >> >> seems necessary, then how about this strawman: > >> >> > >> >> * Someone can be nominated to be a reviewer of an area by sending a > >> >> private email to every reviewer and maintainer who covers a non-strict > >> >> superset of that area. The nomination is approved if it is supported > >> >> by at least two such reviewers or maintainers and if there are no > >> >> objections. People would be given at least a week to respond. > >> >> > >> >> * The process would be the same for maintainers, with the same set of > >> >> addressees, except that there must already be at least one maintainer > >> >> for that area and, in addition to the previous requirements, all such > >> >> maintainers must be in favour. > >> >> > >> >> (So if the area is maintained by one person, the nomination would > >> >> need the support of that maintainer and at least one reviewer of a > >> >> wider area. If the area is maintained by two of more people, they > >> >> would all need to agree.) > >> >> > >> >> The idea with making it private is that it allows for a more honest > >> >> discussion. But the convention could be to have a public discussion > >> >> instead, if that seems better. > >> >> > >> >> Like I say, this would just be a second, alternative route. It would > >> >> still be possible to ask the SC instead. > >> >> > >> >> In case it sounds otherwise, I'm really not trying to pick a fight here. > >> >> I just don't remember this being discussed on-list for a long time, > >> >> so it seemed worth bringing up. (Maybe it has been discussed at the > >> >> Cauldron -- not sure.) > >> >> > >> > > >> > What is a request to the GCC SC preventing? > >> > > >> > The GCC SC already requests the opinion of existing reviewers and > >> > maintainers. > >> > >> Like I say, the idea isn't to replace the existing system, just to > >> provide a second, alternative path. > >> > >> But I suppose the question works both ways: does the SC need to be > >> involved in every decision? There doesn't seem to be a specific need > >> for the SC to act as the gatherer of opinions if the maintainers/reviewers > >> are able to agree on a candidate directly amongst themselves. In cases > >> like those, having the conversation directly would be a lighter-weight > >> and more transparent process (especially if, as Richard suggests, > >> the discussion happens in public). > >> > > > > What is not working with the current system? What is this fixing? > > > > The GCC SC has not been notified of any problems with appointing > > maintainers. > > The SC is (by design, as I understand) opaque to those not in the SC. > How the SC works, and how it makes decisions, are not visible to outsiders. > > There is no defined way of contacting the SC collectively. > You have to approach individual members, ask them to contact the SC on > your behalf, then wait for the SC's decision to be forwarded back to you. > > The problem with this approach is that it's easy for things to get lost. > Last year I nominated someone for a new role and the whole process > took over three months. > > In more detail: I originally sent a nomination to two members of the SC > and asked them to forward it to the wider SC. I didn't hear anything > for a while. But because of the way that the SC works, I didn't expect > to hear anything for a while, so nothing seemed wrong. However, it turned > out that both of the people that I contacted had overlooked the message, > and so in fact nothing had happened. I don't blame them at all for that > -- we're all busy people, and I could so very easily have done the same > thing in their position! But it was only after a couple of months of > waiting that the process proper actually started. > > Of course, I'd asked the nominee beforehand whether they were > interested, so such a long delay didn't give a good impression. > I began to wonder, and I wouldn't blame the person I nominated for > wondering, whether the SC had some doubts about their ability. > It all seemed like a needless worry. > > I have heard another instance of a discussion with the SC suffering > similar hiccups. It's a natural risk when each back-and-forth relies > on a member of the SC acting as an intermediary. > > Of course, there are times when being able to deliberate in secret > is useful. But I'm not sure why appointing reviewers and maintainers > is one of them, if existing existing maintainers/reviewers of the same > area are in clear agreement. Does the SC need to be involved in every > such decision?
My main concern here is that only three SC members are actively participating in GCC development and thus might have an own understanding of a candidates applicability. That doesn't say the SC (or one SC member) might not be able to reach out to appropriate people outside of the SC to gather opinions. It's just the process is whoefully intransparent and in my experience there's often no communication back to the requestor other than that there's eventually a SC announcement of a new maintainer/reviewer. I'm not asking for a SC ticket system, but even that would be an improvement ... Richard. > > Thanks, > Richard
