Re: [gentoo-dev] [PATCH] glep-0067: Add 'proxied' and 'watcher' maint types
On Tue, 06 Aug 2019 18:53:33 +0200 Ulrich Mueller wrote: > Yes, I know how elementary set theory works. :) I had asked for a > concrete _example_ in the outer set but not in the inner. Easy: - You maintain a package that depends on this, and you want to be kept "in the loop" on changes/bugs just in case they necessitate changes of your own - You're maintaining an independent overlay that contains derivative forms of stuff in ::gentoo, and so bugs that affect one likely affect the other Neither of these pass for 'maintainer' because they have no authority with regards to the package in question. So the nested sets basically communicate the level of authority one has, the outer one having the least, and the inner one having an authority that transcends gentoo. pgprdqS1ctaam.pgp Description: OpenPGP digital signature
Re: [gentoo-dev] [PATCH] glep-0067: Add 'proxied' and 'watcher' maint types
> On Tue, 06 Aug 2019, Kent Fredric wrote:
>> Can you give an example of a maintainer in the watcher \ upstream set?
> all maintainers who aren't upstream are just maintainers
> all maintainers who are upstream are just maintainers
> all maintainers are watchers
> So:
> { watcher { maintainer { upstream }}}
Yes, I know how elementary set theory works. :) I had asked for a
concrete _example_ in the outer set but not in the inner.
Ulrich
signature.asc
Description: PGP signature
Re: [gentoo-dev] [PATCH] glep-0067: Add 'proxied' and 'watcher' maint types
On Fri, 2019-08-02 at 22:55 +0200, Michał Górny wrote: > Add two new maintainer types: 'proxied' for proxied maintainers, > and 'watcher' for people who wish to be CC-ed on bugs but are not > maintainers (e.g. upstream developers). > > While at it, clarify that subproject member inheritance does not carry > over lead status to parent projects. > I withdraw this patch. -- Best regards, Michał Górny signature.asc Description: This is a digitally signed message part
Re: [gentoo-dev] [PATCH] glep-0067: Add 'proxied' and 'watcher' maint types
On Tue, 06 Aug 2019 09:45:18 +0200
Ulrich Mueller wrote:
> > All fit within
> > , but not all
> > fits within
>
> Can you give an example of a maintainer in the watcher \ upstream set?
>
> Ulrich
all maintainers who aren't upstream are just maintainers
all maintainers who are upstream are just maintainers
all maintainers are watchers
So:
{ watcher { maintainer { upstream }}}
pgp7E6Q3l4ixT.pgp
Description: OpenPGP digital signature
Re: [gentoo-dev] [PATCH] glep-0067: Add 'proxied' and 'watcher' maint types
> On Tue, 06 Aug 2019, Kent Fredric wrote: > All fit within > , but not all > fits within Can you give an example of a maintainer in the watcher \ upstream set? Ulrich signature.asc Description: PGP signature
Re: [gentoo-dev] [PATCH] glep-0067: Add 'proxied' and 'watcher' maint types
On Sun, 04 Aug 2019 18:49:30 +0200 Ulrich Mueller wrote: > > And yes, I'm talking about real life situation when the only > > in the package left was this 'upstream watcher'. > > I suppose an alternative solution there would be to return to explicit > > logical marking as . > > Many metadata files have that anyway as a comment, which is far from > perfect. So yes, I'd say that explicit is better > than . > > Alternatively, how about calling that type "upstream" instead of > "watcher"? Hmm, actually, maybe what this calls for is a new tag, "", to denote involved entities that aren't maintainers, but need to be CC'd on bugs. e.g.: Therein, a package with no is unmaintained, but people in the CC list still get CC'd, and a package with neither or is a bug. Perhaps even stipulate a 3rd tag, which repoman enforces being present if the count of drops below 1, and indicates that the Assignment on bugzilla should be to maintainer-needed? pgpZgR9OZeQjL.pgp Description: OpenPGP digital signature
Re: [gentoo-dev] [PATCH] glep-0067: Add 'proxied' and 'watcher' maint types
On Sun, 04 Aug 2019 18:49:30 +0200 Ulrich Mueller wrote: > Alternatively, how about calling that type "upstream" instead of > "watcher"? Mostly, because the term "upstream" doesn't communicate any useful information about what it is expected to mean, and, it reduces the usefulness of this field to excluding people who might pass for "watcher" but don't pass for "upstream" There are already "upstream" fields in other parts of metadata.xml, but none of them indicate definitively if upstream should (or shouldn't) be CC'd on literally every bug. All fit within , but not all fits within pgpcs5t86egaB.pgp Description: OpenPGP digital signature
Re: [gentoo-dev] [PATCH] glep-0067: Add 'proxied' and 'watcher' maint types
> On Sat, 03 Aug 2019, Michał Górny wrote: >> Upstream developers can be listed already now in the >> description (per GLEP 68). Should they be listed twice now, only to >> indicate that they are to be CCed on bugs? > This is happening already. I'm not saying it's perfect but I don't see > anyone working on a better solution either. > And yes, I'm talking about real life situation when the only > in the package left was this 'upstream watcher'. > I suppose an alternative solution there would be to return to explicit > logical marking as . Many metadata files have that anyway as a comment, which is far from perfect. So yes, I'd say that explicit is better than . Alternatively, how about calling that type "upstream" instead of "watcher"? Ulrich signature.asc Description: PGP signature
Re: [gentoo-dev] [PATCH] glep-0067: Add 'proxied' and 'watcher' maint types
Hi On 2019/08/03 01:19, Jonas Stein wrote: On 02/08/2019 22.55, Michał Górny wrote: Add two new maintainer types: 'proxied' for proxied maintainers, and 'watcher' for people who wish to be CC-ed on bugs but are not maintainers (e.g. upstream developers). Can't we solve this simply in the bug tracker? The monitor setting of a user does not belong into the tree. The upstream maintainer and all other "watchers" have no write access to the tree so they will consume manpower in adding and removing their contacts to packages. The perfect solution would be that any user can add a watch filter to my-cat/mypkg in the bugtracker. Between 2018-01-01 and 2018-12-31 we received and assigned 31280 bugs. I am no fan of the descriptions in the form "please CC: If the bug is about x but not y and the moon is in the third house of the lion" This consumes extra time for every assignment and prevents automagic assignment in future. We should rather keep it simple instead of extending the options. I agree. I was just thinking "I'd like to watch certain packages, be aware of commits et al, but not get involved in ongoing stuff" too. The rule should be simple: CC me on any bugs to this list of packages. I can then unsub myself from individual bugs I don't care about. Then there are packages that have changed in the past unexpectedly and bit me (badly sometimes), so on those I'd like to know about any commit going in. Guess I can set up a git pull + filter myself there, but if this is done upstream then it can work for everyone, and not waste duplication effort. Kind Regards, Jaco
Re: [gentoo-dev] [PATCH] glep-0067: Add 'proxied' and 'watcher' maint types
On Sat, 3 Aug 2019 01:20:34 +0200 Jonas Stein wrote: > I am no fan of the descriptions in the form "please CC: If the bug is > about x but not y and the moon is in the third house of the lion" You could stipulate that in order to be added as a "watcher" in metadata.xml, you must agree to accept to be CC'd on all and any bugs about the package, and handle filtering what is, and what is not relevant, yourself. You're free to un-CC yourself once added, after all. If you don't want the spam, don't volunteer to be a watcher. pgpzCD_z4EEKZ.pgp Description: OpenPGP digital signature
Re: [gentoo-dev] [PATCH] glep-0067: Add 'proxied' and 'watcher' maint types
On Sat, 2019-08-03 at 01:20 +0200, Jonas Stein wrote: > On 02/08/2019 22.55, Michał Górny wrote: > > Add two new maintainer types: 'proxied' for proxied maintainers, > > and 'watcher' for people who wish to be CC-ed on bugs but are not > > maintainers (e.g. upstream developers). > > Can't we solve this simply in the bug tracker? The monitor setting of a > user does not belong into the tree. > > > The upstream maintainer and all other "watchers" have no write access to > the tree so they will consume manpower in adding and removing their > contacts to packages. I agree, and I'm actually afraid this might happen long-term. However, I don't think it's a real problem right now since the feature is used rarely. > > The perfect solution would be that any user can add a watch filter to > my-cat/mypkg in the bugtracker. Sure. But AFAIK Bugzilla does not have such a function. > > Between 2018-01-01 and 2018-12-31 we received and assigned 31280 bugs. > I am no fan of the descriptions in the form "please CC: If the bug is > about x but not y and the moon is in the third house of the lion" > > This consumes extra time for every assignment and prevents automagic > assignment in future. We should rather keep it simple instead of > extending the options. This is irrelevant to the topic at hand. If explicit type="" for this is added, you don't have to do anything special. Everything can be scripted correctly. > > > > +There are four defined maintainer types: > > + > > +person > > + Indicates a maintainer that is not a project, and has commit access > > + to the repository. > > + > > +proxied > > + Indicates a proxied maintainer, i.e. a maintainer that does not have > > + direct commit access and needs a proxy to commit the changes. > > + > > +watcher > > + Indicates a non-maintainer that wishes to be CC-ed on bug reports, > > + e.g. a upstream developer. > > + > > +project > > + Indicates a maintainer that is a project defined in ``projects.xml``. > > I think different "objects" should not be stored in one variable. > "Person" or "Proxied" means here write_access=TRUE/FALSE > "Project" means a group of developers > "Watcher" means let Bugzilla send notifications I really do want to avoid having exponential growth of fields. For example, a 'project' can't really be a proxied maintainer (it can include proxied maintainers though). We'd end up having a lot of fields with special relations and exclusive blocks. > "Person": We should name people without write access persons too... at > least in public ;-) > > > We should not introduce "watchers" as long our upstream fields in the > metadata are out of date or not set at all. > > > > +Previously, proxied maintainers were distinguished by not having a Gentoo > > e-mail > > +address. However, nowadays we have developers without direct Gentoo > > repository > > +commit access as maintainers. > > (formerly known as staffers) > these few can use their non gentoo mail, if they do not want to do the quiz. > But I think, if someone has done the staffer quiz and is very active on > the ebuilds it makes most sense to go for the write access. There is one developer who does not accept GLEP 76 and therefore commits via proxy. -- Best regards, Michał Górny signature.asc Description: This is a digitally signed message part
Re: [gentoo-dev] [PATCH] glep-0067: Add 'proxied' and 'watcher' maint types
On Sat, 2019-08-03 at 06:40 +0200, Ulrich Mueller wrote: > > > > > > On Fri, 02 Aug 2019, Michał Górny wrote: > > Add two new maintainer types: 'proxied' for proxied maintainers, > > and 'watcher' for people who wish to be CC-ed on bugs but are not > > maintainers (e.g. upstream developers). > > Upstream developers can be listed already now in the > description (per GLEP 68). Should they be listed twice now, only to > indicate that they are to be CCed on bugs? This is happening already. I'm not saying it's perfect but I don't see anyone working on a better solution either. And yes, I'm talking about real life situation when the only in the package left was this 'upstream watcher'. I suppose an alternative solution there would be to return to explicit logical marking as . > On the other hand, random non-upstream people that don't have any > defined function related to the package aren't maintainers. So they > don't belong in metadata at all, IMHO. > I agree. -- Best regards, Michał Górny signature.asc Description: This is a digitally signed message part
Re: [gentoo-dev] [PATCH] glep-0067: Add 'proxied' and 'watcher' maint types
> On Fri, 02 Aug 2019, Michał Górny wrote: > Add two new maintainer types: 'proxied' for proxied maintainers, > and 'watcher' for people who wish to be CC-ed on bugs but are not > maintainers (e.g. upstream developers). Upstream developers can be listed already now in the description (per GLEP 68). Should they be listed twice now, only to indicate that they are to be CCed on bugs? On the other hand, random non-upstream people that don't have any defined function related to the package aren't maintainers. So they don't belong in metadata at all, IMHO. Ulrich signature.asc Description: PGP signature
Re: [gentoo-dev] [PATCH] glep-0067: Add 'proxied' and 'watcher' maint types
Hi, Pleased to meet you i am proxy-maintainer in gentoo :) What i see here i got a new title something like proxied or without-commit-acces-proxy-maintainer or super-proxy-maintainer still not clear for me. As a very active proxy-maintainer in gentoo for a year i think it is not important how you are splitting things in metadata like person, proxied, non-proxied, robot, cyborg .. Also it is not important who has @gentoo mail. I am trying to manage many package in gentoo (without paycheck) and i can say that proxy-maint has lost its function because of super-busy gentoo-devs. Sorry but if you are super-busy give it up. If you have no time to test PRs or your own packages then simply retire. If you don't have time to install manpages with use flag then retire. Being only 7/24 online on IRC doesn't mean you are a dev. If you scary about new packages or big PRs then again give up. I wrote 900 line openblas switch script 5 month ago for gentoo reference-blas-lapack set .This PR waited 5 months and nobody cared it and an other dev merged his own openblas PR with new switch framework without informing me. If a dev not checking active PRs before opening a new one there is no word to say. A simple example to explain a situation what is happening on gentoo currently. Anybody checked BIND package version in tree ? It is 9.12.3_P4 [1] EOL as of May 2019. This is the one of the core package any linux distro. I updated bind ebuild then upgraded my DNS server but i didn't opened any PR because i know you guys so-busy. Anyway thank you for your great support. Best. ~Hasan 2019-08-03 03:26, Aaron Bauman yazmış: On Sat, Aug 03, 2019 at 01:20:34AM +0200, Jonas Stein wrote: On 02/08/2019 22.55, Michał Górny wrote: Add two new maintainer types: 'proxied' for proxied maintainers, and 'watcher' for people who wish to be CC-ed on bugs but are not maintainers (e.g. upstream developers). Can't we solve this simply in the bug tracker? The monitor setting of a user does not belong into the tree. I would disagree with this. The benefits of it being the metadata is just that... it is metadata. Hence, QA checks can logically determine the state of maintainership on a given package. This is one of the reasons metadata.xml was standardized years ago. The upstream maintainer and all other "watchers" have no write access to the tree so they will consume manpower in adding and removing their contacts to packages. This should become a repetitive task unless the package is continuously being turned over to other maintainers. Once and done is the general rule. The perfect solution would be that any user can add a watch filter to my-cat/mypkg in the bugtracker. This assumes that those reporting bugs are inputing proper information. See my first comment regarding the metadata.xml standards. Between 2018-01-01 and 2018-12-31 we received and assigned 31280 bugs. I am no fan of the descriptions in the form "please CC: If the bug is about x but not y and the moon is in the third house of the lion" This is a good point. I am not sure how many packages have this type of information in the metadata, but it is not a good place for it. This consumes extra time for every assignment and prevents automagic assignment in future. We should rather keep it simple instead of extending the options. I would think this *would* help us do automagic assignements in the future, but more accurately. Links: -- [1] https://gitweb.gentoo.org/repo/gentoo.git/tree/net-dns/bind/bind-9.12.3_p4.ebuild signature.asc Description: OpenPGP digital signature
Re: [gentoo-dev] [PATCH] glep-0067: Add 'proxied' and 'watcher' maint types
On Sat, Aug 03, 2019 at 01:20:34AM +0200, Jonas Stein wrote: > On 02/08/2019 22.55, Michał Górny wrote: > > Add two new maintainer types: 'proxied' for proxied maintainers, > > and 'watcher' for people who wish to be CC-ed on bugs but are not > > maintainers (e.g. upstream developers). > > Can't we solve this simply in the bug tracker? The monitor setting of a > user does not belong into the tree. I would disagree with this. The benefits of it being the metadata is just that... it is metadata. Hence, QA checks can logically determine the state of maintainership on a given package. This is one of the reasons metadata.xml was standardized years ago. > > The upstream maintainer and all other "watchers" have no write access to > the tree so they will consume manpower in adding and removing their > contacts to packages. > This should become a repetitive task unless the package is continuously being turned over to other maintainers. Once and done is the general rule. > The perfect solution would be that any user can add a watch filter to > my-cat/mypkg in the bugtracker. > This assumes that those reporting bugs are inputing proper information. See my first comment regarding the metadata.xml standards. > Between 2018-01-01 and 2018-12-31 we received and assigned 31280 bugs. > I am no fan of the descriptions in the form "please CC: If the bug is > about x but not y and the moon is in the third house of the lion" > This is a good point. I am not sure how many packages have this type of information in the metadata, but it is not a good place for it. > This consumes extra time for every assignment and prevents automagic > assignment in future. We should rather keep it simple instead of > extending the options. > I would think this *would* help us do automagic assignements in the future, but more accurately. -- Cheers, Aaron signature.asc Description: PGP signature
Re: [gentoo-dev] [PATCH] glep-0067: Add 'proxied' and 'watcher' maint types
On 02/08/2019 22.55, Michał Górny wrote: > Add two new maintainer types: 'proxied' for proxied maintainers, > and 'watcher' for people who wish to be CC-ed on bugs but are not > maintainers (e.g. upstream developers). Can't we solve this simply in the bug tracker? The monitor setting of a user does not belong into the tree. The upstream maintainer and all other "watchers" have no write access to the tree so they will consume manpower in adding and removing their contacts to packages. The perfect solution would be that any user can add a watch filter to my-cat/mypkg in the bugtracker. Between 2018-01-01 and 2018-12-31 we received and assigned 31280 bugs. I am no fan of the descriptions in the form "please CC: If the bug is about x but not y and the moon is in the third house of the lion" This consumes extra time for every assignment and prevents automagic assignment in future. We should rather keep it simple instead of extending the options. > +There are four defined maintainer types: > + > +person > + Indicates a maintainer that is not a project, and has commit access > + to the repository. > + > +proxied > + Indicates a proxied maintainer, i.e. a maintainer that does not have > + direct commit access and needs a proxy to commit the changes. > + > +watcher > + Indicates a non-maintainer that wishes to be CC-ed on bug reports, > + e.g. a upstream developer. > + > +project > + Indicates a maintainer that is a project defined in ``projects.xml``. I think different "objects" should not be stored in one variable. "Person" or "Proxied" means here write_access=TRUE/FALSE "Project" means a group of developers "Watcher" means let Bugzilla send notifications "Person": We should name people without write access persons too... at least in public ;-) We should not introduce "watchers" as long our upstream fields in the metadata are out of date or not set at all. > +Previously, proxied maintainers were distinguished by not having a Gentoo > e-mail > +address. However, nowadays we have developers without direct Gentoo > repository > +commit access as maintainers. (formerly known as staffers) these few can use their non gentoo mail, if they do not want to do the quiz. But I think, if someone has done the staffer quiz and is very active on the ebuilds it makes most sense to go for the write access. > +There is also no reason to assume that > +in the future we would not have developers using non-Gentoo e-mail addresses. > +Adding explicit notation for proxied maintainers resolves that problem. This would make it impossible to see outside of Gentoo, who is who. I think users have a slightly higher trust in a maintainer with f...@gentoo.org than superhero1...@maildrop.cc -- Best, Jonas signature.asc Description: OpenPGP digital signature
Re: [gentoo-dev] [PATCH] glep-0067: Add 'proxied' and 'watcher' maint types
On 02/08/2019 22.55, Michał Górny wrote: > Add two new maintainer types: 'proxied' for proxied maintainers, > and 'watcher' for people who wish to be CC-ed on bugs but are not > maintainers (e.g. upstream developers). Can't we solve this simply in the bug tracker? The monitor setting of a user does not belong into the tree. The upstream maintainer and all other "watchers" have no write access to the tree so they will consume manpower in adding and removing their contacts to packages. The perfect solution would be that any user can add a watch filter to my-cat/mypkg in the bugtracker. Between 2018-01-01 and 2018-12-31 we received and assigned 31280 bugs. I am no fan of the descriptions in the form "please CC: If the bug is about x but not y and the moon is in the third house of the lion" This consumes extra time for every assignment and prevents automagic assignment in future. We should rather keep it simple instead of extending the options. > +There are four defined maintainer types: > + > +person > + Indicates a maintainer that is not a project, and has commit access > + to the repository. > + > +proxied > + Indicates a proxied maintainer, i.e. a maintainer that does not have > + direct commit access and needs a proxy to commit the changes. > + > +watcher > + Indicates a non-maintainer that wishes to be CC-ed on bug reports, > + e.g. a upstream developer. > + > +project > + Indicates a maintainer that is a project defined in ``projects.xml``. I think different "objects" should not be stored in one variable. "Person" or "Proxied" means here write_access=TRUE/FALSE "Project" means a group of developers "Watcher" means let Bugzilla send notifications "Person": We should name people without write access persons too... at least in public ;-) We should not introduce "watchers" as long our upstream fields in the metadata are out of date or not set at all. > +Previously, proxied maintainers were distinguished by not having a Gentoo > e-mail > +address. However, nowadays we have developers without direct Gentoo > repository > +commit access as maintainers. (formerly known as staffers) these few can use their non gentoo mail, if they do not want to do the quiz. But I think, if someone has done the staffer quiz and is very active on the ebuilds it makes most sense to go for the write access. > +There is also no reason to assume that > +in the future we would not have developers using non-Gentoo e-mail addresses. > +Adding explicit notation for proxied maintainers resolves that problem. This would make it impossible to see outside of Gentoo, who is who. I think users have a slightly higher trust in a maintainer with f...@gentoo.org than superhero1...@maildrop.cc signature.asc Description: OpenPGP digital signature
