Re: [oi-dev] Where should SPARC go?
Gary, 1. Testing a newer compile of the current illumos-gate snapshot seems reasonable to ensure everything still works with your SPARC-related changes. Ref: https://github.com/illumos/illumos-gate/commit/f52943a93040563107b95bccb9db87d9971ef47d I'll review your SPARC-related tickets, if you have any porting issues with recent oi-userland on SPARC - did you provide a list as well? ~ Ken On Tuesday, November 26, 2019, 6:47:33 AM PST, Gary Mills wrote: On Sun, Nov 24, 2019 at 08:40:58PM +, Peter Tribble wrote: > On Fri, Nov 22, 2019 at 8:43 PM Gary Mills <[1]gary_mi...@fastmail.fm> > wrote: > > I have no doubt that SPARC owners are in the minority.� My estimate > is > about 12 users and 6 developers. > > That's probably slightly low in terms of users. There are currently > about 6 active > users of Tribblix for SPARC (other than myself). (For reference, there > are 50-100 > active users of Tribblix on x86.) I suspect the number would increase > if illumos on > SPARC were more mature. That's probably true. It's still a small minority, of course. > That's partly because illumos already has SPARC support built in: > Peter only fixes things in illumos that are broken.� > > It's not as if there's any hardware development going on that affects > the > supported SPARC platforms. (Yes, it would be nice to support more > current > hardware, but what we have is a static target.) Much of the cleanup > work that > I'm (very very slowly) doing is to try and reduce the blast radius of > changes > from x86. Yes, I agree with that policy. Ordinary people have to buy SPARC machines on the used market. The largest SPARC machines are too expensive. The oldest ones are too slow to be useful. It's only the mid-range that we need to support. Even those are large, heavy, and power-hungry. > Full independance is a good way to cripple SPARC development.� OI > source archives, Makefiles, and package manifests are used to build > and publish OI SPARC packages, generally with no change.� It's OI > all > the way.� The only thing that will change is the list of components > to > build. > > I wouldn't have expected full independence. I would have thought you > want > your own fork of oi-userland, though, so you can make whatever local > changes > necessary. There won't be any local changes. At least, that is my intention. > One thing I'm finding is that the chance of something > building on > SPARC is going down over time, so I'm having to hold certain packages > on > SPARC at older revisions. So far, I haven't run into that problem with OI. My intention is to build OI packages without any changes. Most of the time, this plan is successful. -- -Gary Mills- -refurb- -Winnipeg, Manitoba, Canada- ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev
Re: [oi-dev] Where should SPARC go?
On Sun, Nov 24, 2019 at 08:40:58PM +, Peter Tribble wrote: >On Fri, Nov 22, 2019 at 8:43 PM Gary Mills <[1]gary_mi...@fastmail.fm> >wrote: > > I have no doubt that SPARC owners are in the minority.� My estimate > is > about 12 users and 6 developers. > >That's probably slightly low in terms of users. There are currently >about 6 active >users of Tribblix for SPARC (other than myself). (For reference, there >are 50-100 >active users of Tribblix on x86.) I suspect the number would increase >if illumos on >SPARC were more mature. That's probably true. It's still a small minority, of course. > That's partly because illumos already has SPARC support built in: > Peter only fixes things in illumos that are broken.� > >It's not as if there's any hardware development going on that affects >the >supported SPARC platforms. (Yes, it would be nice to support more >current >hardware, but what we have is a static target.) Much of the cleanup >work that >I'm (very very slowly) doing is to try and reduce the blast radius of >changes >from x86. Yes, I agree with that policy. Ordinary people have to buy SPARC machines on the used market. The largest SPARC machines are too expensive. The oldest ones are too slow to be useful. It's only the mid-range that we need to support. Even those are large, heavy, and power-hungry. > Full independance is a good way to cripple SPARC development.� OI > source archives, Makefiles, and package manifests are used to build > and publish OI SPARC packages, generally with no change.� It's OI > all > the way.� The only thing that will change is the list of components > to > build. > >I wouldn't have expected full independence. I would have thought you >want >your own fork of oi-userland, though, so you can make whatever local >changes >necessary. There won't be any local changes. At least, that is my intention. >One thing I'm finding is that the chance of something >building on >SPARC is going down over time, so I'm having to hold certain packages >on >SPARC at older revisions. So far, I haven't run into that problem with OI. My intention is to build OI packages without any changes. Most of the time, this plan is successful. -- -Gary Mills--refurb--Winnipeg, Manitoba, Canada- ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev
Re: [oi-dev] Where should SPARC go?
On 11/22/19 09:43 PM, Gary Mills wrote: On Thu, Nov 21, 2019 at 10:56:11PM +0100, Michal Nowak wrote: On 11/17/19 10:39 PM, Gary Mills wrote: I have a series of questions to ask the members of this mailing list. I have information to share with you as well. I already have partial answers to most of the questions, but I'd still like to hear from you. [...] There were 7 contributors to the OpenIndiana/oi-userland repo (which is the bulk of OpenIndiana development) with more than 10 commits for the 2019.10 release. I don't know how many OI users (be it long-term or accidental ones) are out there for the same period of time but I guess it's in low hundreds. I have no doubt that SPARC owners are in the minority. My estimate is about 12 users and 6 developers. Although I consider Gary's OI on SPARC an interesting achievement, I doubt there are enough potential users and developers to sustain any illumos SPARC distro of the same (slow) pace of hipster development. E.g. looking at SPARC-related work in illumos, it's only Peter who contributes significantly to this area (mostly by removing entirely obsolete platforms and by ensuring illumos actually builds - and that's, I am afraid, revealing). That's partly because illumos already has SPARC support built in: Peter only fixes things in illumos that are broken. I've submitted 14 bug reports, with patches, to the OI project. I'm aware that the patches need to be turned into PRs before OI developers will consider them. Most of the bug reports are not SPARC-specific. I did not need to submit any bug reports to the illumos project. That built-in support is for > 10 years old platforms, for me that is indicative about interest in this platform. Given that illumos SPARC distros were not closely following illumos development, years of testing stuff which gets added is missing for SPARC. Given the tiny resources OI project has and can offer, I wish OI and "OI"-for-SPARC were independent and related to each other in the same way OI currently relates to OmniOS CE and pkgsrc, that is by sharing knowledge, patches. Full independance is a good way to cripple SPARC development. OI source archives, Makefiles, and package manifests are used to build and publish OI SPARC packages, generally with no change. It's OI all the way. The only thing that will change is the list of components to build. The package names are the same. The package repository does not have to be hosted by OI, of course. I've had a few offers of a build server with jenkins. It will clone oi-userland periodically, and build packages from it. Only the binary packages will be different from the x86 version. The whole thing could be fully automated and will track OI pretty closely. We are not there yet, but we will be. Whatever the terminology on the projects relation, what I am really interested in is the OI amd64 development not being adversely affected. Although I am not excited about it, I don't mind SPARC changes in the OpenIndiana GitHub projects (Makefiles, patches, .p5m files, ...), provided that OI does not guarantee anything for SPARC, including: * we update a component which fails to build on SPARC, we don't retract, * updated component's SPARC patch fails to apply and it's too complicated to fix it, we remove it for the time being and re-apply when SPARC developers come up with a fixed patch, * we don't test that changes build the less work on SPARC I don't think that's a great experience for SPARC re-builders from OpenIndiana/oi-userland. Michal ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev
