Where is the science overlay? I can't find it no more with layman -L nor eselect repository list
It has been removed because it's broken and its maintainers don't care.
(In reply to Michał Górny from comment #1) > It has been removed because it's broken Would you mind to substantiate a bit what exactly is broken right now? Even though we are currently discussing within the science project to phase out the overlay at some point in the near future, I'd say it would nevertheless be a bit more appropriate for us to call the shots on that one. Would you mind relisting the overlay please?
commit 924f038b69dfc701e3878aec5cfc5f4d1610b6ec (HEAD -> master) Author: Matthias Maier <tamiko@43-1.org> Date: Tue Sep 3 15:04:13 2019 -0500 repositories: put the science overlay back in place Signed-off-by: Matthias Maier <tamiko@43-1.org>
(In reply to Matthias Maier from comment #3) > commit 924f038b69dfc701e3878aec5cfc5f4d1610b6ec (HEAD -> master) > Author: Matthias Maier <tamiko@43-1.org> > Date: Tue Sep 3 15:04:13 2019 -0500 > > repositories: put the science overlay back in place > > Signed-off-by: Matthias Maier <tamiko@43-1.org> Did you push that? I don't see that commit anywhere.
(In reply to Michał Górny from comment #4) > (In reply to Matthias Maier from comment #3) > > commit 924f038b69dfc701e3878aec5cfc5f4d1610b6ec (HEAD -> master) > > Author: Matthias Maier <tamiko@43-1.org> ^~~~~~~~~~~~~~~ > > Date: Tue Sep 3 15:04:13 2019 -0500 > > > > repositories: put the science overlay back in place > > > > Signed-off-by: Matthias Maier <tamiko@43-1.org> ^~~~~~~~~~~~~~~ > > Did you push that? I don't see that commit anywhere. No, I did not. (I didn't notice the error message.) As soap and I have reached the conclusion that we should actually simply retire the overlay, I will also not push that commit at the moment.
For the record, I have sent the following e-mail to the sci@gentoo.org mail alias: Dear all, As some of you have noticed the science overlay got removed from the overlay repositories list a couple of days ago. This is due to multiple unresolved QA issues as can be seen here [1]. Looking at this huge list of issues and considering the fact that almost none of the sci project members (i.e. Gentoo developers) is active in the overlay any more to maintain a certain level of QA standards, David (soap) and I decided to propose the only viable option - namely to retire the overlay. It is no coincidence that the decline of the Science Overlay started when pull requests (and git) became much more usable for user contributions to the main repository. Thus, we suggest the following move forward: - We will retire (i.e. delete) the overlay in a 30 day timeframe. - This should allow ample time to * migrate packages that are fit for a general audience to the main repository (either maintained by sci@gentoo, or by sci@gentoo + proxy-maint@gentoo + user maintainer) - provided the packages meet the standards for ::gentoo * discuss the creation of one (or more) overlays, e.g. "cluster", for stuff dedicated to a single purpose with the installment of a number of *mandatory* QA policies: - all but latest EAPI banned - no creation of custom eclasses - commits that trigger any repoman warning (as seen in [1]) are getting reverted on a semi-automatic manner. Best, Matthias [1] https://travis-ci.org/gentoo/sci/builds/579626517
Hi, Matthias, I agree with this decision. But we have many active overlay users. I suggest we invite them to express their opinions, too. So, it makes more sense to circulate an email to gentoo-science mailing list and hold on with the "30 days" deadline before a conclusion involving more interested parties is drawn. Benda
Just mentioning, I have adopted sci-mathematics/lie in the sage-on-gentoo overlay [and gave it a serious clean] as it is an optional dependency of sage. For the same reasons, I may be interested in adopting gambit kash-bin I already have better/more recent ebuilds for givaro gap giac all currently in the science overlay.
Maybe it's time to give https://bugs.gentoo.org/201321 some love
User, but not maintainer, of the science overlay here. I update a little less than weekly, and I sometimes submit bug reports here. I'm a Ph.D. scientist with over 10,000 citations to my work in bioinformatics and structural biology. I actively develop code every week. Removing the science overlay is an "Oh hell no!" for me. Delisting it as you have done is already the equivalent of jabbing your thumb in my eye. Deleting it would be like lighting a library of books on fire because the bindings are frayed and some volumes haven't been checked out in a few years. I chose Gentoo as my developer-desktop distro largely because of packages in the science overlay. Though it's been a bit quiet lately, seeing new packages appear in the overlay is an important news source for me. If the decision sticks to delist it, that's a sign I should reconsider my selection of Gentoo on my developer desktop and put my effort into Arch or Clear Linux (both of which I use and like in other environments) or to concentrated on Docker packaging. I'll be a bit sad because Gentoo is still about the richest distro in terms of scientific packages and ebuilds are still a great source of shared wisdom on how to install a package. But deleting the science overlap will be a huge sign that it's time to move on. eix shows that I have 128 packages installed from the science overlay, mostly in sci-biology and dev-python. I have roughly the same number of packages installed from my private repository, many of which are based on science repository entries with fixes or updates applied. I'm sorry to break this to you, but having some packages that don't pass the QA checks is a normal state of affairs in the life cycle of scientific software. If Gentoo's QA process can't handle some broken packages, that's a problem with the QA process rather than a reason to kick out a whole repository. Scientific software development is generally written by a single author with middling-to-poor software engineering skills and an urgent need to finish a project. There is rarely funding or mandate to maintain scientific software, even if it proves highly useful and wildly popular for a time. Later would-be users with software skills have to decide whether it's better to fix the problems or to re-implement from scratch, possibly with better algorithms. Being able to find an ebuild that once worked is a far better starting point than having nothing at all. That's not to say that the science overlay couldn't use some work and additional involvement. I'm guilty as the next data scientist on that one. As for splitting, I'd say "no" unless repositories could be made hierarchical because nearly all the science applications share dependencies on things like linear algebra and, increasingly, machine learning code running on a GPU. Until a few months ago, the main Gentoo repository implementation of BLAS was pretty broken, and I'd assert the model is still wrong for things like MKL with both compile-time and run-time dependencies. Please re-list the science overlay.
(In reply to Joel Berendzen from comment #10) > User, but not maintainer, of the science overlay here. I update a little > less than weekly, and I sometimes submit bug reports here. I'm a Ph.D. > scientist with over 10,000 citations to my work in bioinformatics and > structural biology. I actively develop code every week. > > Removing the science overlay is an "Oh hell no!" for me. Delisting it as you > have done is already the equivalent of jabbing your thumb in my eye. > Deleting it would be like lighting a library of books on fire because the > bindings are frayed and some volumes haven't been checked out in a few years. > > I chose Gentoo as my developer-desktop distro largely because of packages in > the science overlay. Though it's been a bit quiet lately, seeing new > packages appear in the overlay is an important news source for me. If the > decision sticks to delist it, that's a sign I should reconsider my selection > of Gentoo on my developer desktop and put my effort into Arch or Clear Linux > (both of which I use and like in other environments) or to concentrated on > Docker packaging. I'll be a bit sad because Gentoo is still about the > richest distro in terms of scientific packages and ebuilds are still a great > source of shared wisdom on how to install a package. But deleting the > science overlap will be a huge sign that it's time to move on. > > eix shows that I have 128 packages installed from the science overlay, > mostly in sci-biology and dev-python. I have roughly the same number of > packages installed from my private repository, many of which are based on > science repository entries with fixes or updates applied. > > I'm sorry to break this to you, but having some packages that don't pass the > QA checks is a normal state of affairs in the life cycle of scientific > software. If Gentoo's QA process can't handle some broken packages, that's > a problem with the QA process rather than a reason to kick out a whole > repository. > > Scientific software development is generally written by a single author with > middling-to-poor software engineering skills and an urgent need to finish a > project. There is rarely funding or mandate to maintain scientific > software, even if it proves highly useful and wildly popular for a time. > Later would-be users with software skills have to decide whether it's better > to fix the problems or to re-implement from scratch, possibly with better > algorithms. Being able to find an ebuild that once worked is a far better > starting point than having nothing at all. > > That's not to say that the science overlay couldn't use some work and > additional involvement. I'm guilty as the next data scientist on that one. > As for splitting, I'd say "no" unless repositories could be made > hierarchical because nearly all the science applications share dependencies > on things like linear algebra and, increasingly, machine learning code > running on a GPU. Until a few months ago, the main Gentoo repository > implementation of BLAS was pretty broken, and I'd assert the model is still > wrong for things like MKL with both compile-time and run-time dependencies. > > Please re-list the science overlay. That's a very good input to the debate :) That being said, and I am a former physicist with my own list of paper reconverted as a "computing specialist" to help other scientists with their computing problems. At all levels from code to getting some compute hardware. And yes the code is ugly and please at least put it in a public repo and please at least tag the version you use when you publish that paper so that we can go back and checkout that particular version easily for reproducibility. But it can be so ugly that it falls in the category that has a gentoo ebuild maintainer I qualify as "do not want to be packaged". If you are working at my institution I can help with that. If I have an interest in the software and want to see it available in Gentoo, I can help with that too. With that being said a lot of the QA issues have to do with the way the ebuild is written rather than the poor quality of the software [yes it can definitely be a factor but not quite as big as you think]. There are packaging QA and installing QA, while we don't like either most the automated tests above are about the packaging rather than the quality of the code. It all comes down to the lack of man power and to some extent training from the people writing the ebuilds. Should we have a sandpit with low quality ebuilds that people find useful? Possibly. Should it be marked as such? You bet. If you want the science overlay to be that place that may mean a change of policy about what is listed in layman and possibly some gradation of the stuff available. One thing I like about Gentoo is that the packaging is rather "scientifically" split into the phases that you need to consider. And once you have grasped those concepts the barrier to entry for writing a new ebuild is very very low [that's a good thing]. Getting to be able to write an ebuild that pass packaging QA is a higher bar [that's not as good and that's the problem we have right now, we are raising the bar for layman inclusion].
I think losing active users to Gentoo Science Overlay means too much for us. There is no action on the mailing list for further discussion. jlec, would you please call for a Gentoo Science IRC meeting to discuss the fate of science overlay?
All, as we haven't moved in any direction so far, I have reverted the delisting of the overlay by the QA team temporarily. commit 549065020c70efe251a7909755092b1baa3af290 (HEAD -> master, origin/master, origin/HEAD) Author: Matthias Maier <tamiko@gentoo.org> Date: Mon Sep 23 21:06:30 2019 -0500 repositories: readd the science overlay Signed-off-by: Matthias Maier <tamiko@gentoo.org> (In reply to Joel Berendzen from comment #10) > [...] Delisting it as you have done is already the equivalent of jabbing your thumb in my eye. *ouch* That mus hurt :-) > That's not to say that the science overlay couldn't use some work and additional involvement. I'm guilty as the next data scientist on that one. Regarding this point I'll write you an e-mail sometime later this week, if you don't mind. (Sorry for the short reply here, I am quite busy with visitors right now.)
