| Summary: | profiles/releases/23.0/make.defaults: drop or relocate USE=lzma | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Michael Orlitzky <mjo> |
| Component: | Profiles | Assignee: | Andreas K. Hüttel <dilfridge> |
| Status: | RESOLVED FIXED | ||
| Severity: | normal | CC: | bertrand, chris, council, dilfridge, flow, hydrapolic, kripton, kuraga333, mgorny, mjo, pacho |
| Priority: | Normal | ||
| Version: | unspecified | ||
| Hardware: | All | ||
| OS: | Linux | ||
| See Also: | https://bugs.gentoo.org/show_bug.cgi?id=928933 | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
If you're looking for consensus I've been following the threads on gentoo-dev and would agree with this change to the profile. There are other, more sensible, places to put this as a default directly in the packages where it's important. The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=bbe2b57efbc894c3a06c283927ae245d0d1b6454 commit bbe2b57efbc894c3a06c283927ae245d0d1b6454 Author: Andreas K. Hüttel <dilfridge@gentoo.org> AuthorDate: 2024-04-09 21:44:36 +0000 Commit: Andreas K. Hüttel <dilfridge@gentoo.org> CommitDate: 2024-04-09 21:48:17 +0000 profiles, 23.0: Undo USE="lzma zstd", except for toolchain where zstd makes sense As discussed on gentoo-dev and in toolchain / base-system teams. Closes: https://bugs.gentoo.org/928932 Closes: https://bugs.gentoo.org/928933 Signed-off-by: Andreas K. Hüttel <dilfridge@gentoo.org> profiles/releases/23.0/make.defaults | 9 +++------ profiles/releases/23.0/package.use | 10 ++++++++++ 2 files changed, 13 insertions(+), 6 deletions(-) I disagree with making such a change lightly, based on "consensus" that apparently seems to be built by a single Gentoo developer + 3 people who contribute posts to mailing lists. The timeline is particularly bad because: 1. We've enabled it by default. 2. We told all people to switch profiles and rebuild everything. 3. We pull the carpet from under their feet. I can pretty much imagine a scenario where users would suddenly be unable to access their data, because a tool built with USE=lzma started using LZMA compression in them. This is basically opinionated minority sabotaging user systems. (In reply to Michał Górny from comment #3) > I disagree with making such a change lightly, based on "consensus" that > apparently seems to be built by a single Gentoo developer + 3 people who > contribute posts to mailing lists. > > The timeline is particularly bad because: > > 1. We've enabled it by default. > > 2. We told all people to switch profiles and rebuild everything. > > 3. We pull the carpet from under their feet. > > I can pretty much imagine a scenario where users would suddenly be unable to > access their data, because a tool built with USE=lzma started using LZMA > compression in them. > > This is basically opinionated minority sabotaging user systems. If you don't reply to toolchain or base-system pings, your fault. (In reply to Andreas K. Hüttel from comment #4) > If you don't reply to toolchain or base-system pings, your fault. So you're saying that the members of these two teams only get a voice in Gentoo these days? Good to know. It was discussed on IRC in a bunch of places as well where other developers agreed with the change and importantly, nobody spoke up in favour of the status quo. dilfridge is saying he also then pinged core projects just in case to give people a chance to scream before doing it. (In reply to Sam James from comment #6) > It was discussed on IRC in a bunch of places as well where other developers > agreed with the change and importantly, nobody spoke up in favour of the > status quo. Well, I guess I didn't explicitly say I do support "status quo", just indicated that changing it right now could break user systems and cause interoperability issues. My bad. (In reply to Michał Górny from comment #3) > > I can pretty much imagine a scenario where users would suddenly be unable to > access their data, because a tool built with USE=lzma started using LZMA > compression in them. A valid concern, but: 1. Unlike the other way around, it's easy to change by putting USE=lzma in make.conf or in a child profile. 2. Is still hypothetical unless there's an example. 3. If you believe that there are downsides to having the defaults in high-level profiles, and ignoring all else, is motivation to remove those defaults as soon as possible, so that their removal affects as few people as possible. I'm not trying to break user systems or get in the way of people who want the batteries included. There are however a lot of people who use Gentoo _because_ it lets you turn build-time features off. Having the defaults live somewhere else is a compromise that everyone can be happy with. (Thanks for the fast response on this.) |
As discussed on gentoo-dev: This default doesn't actually solve the stated problem, and setting it in a high-level profile causes new ones for users who want it disabled. The obvious solution to revert to the status quo is to set USE="-lzma", but that has the dangerous side-effect of overriding IUSE defaults in packages where they are important. For example, sys-apps/kmod uses +lzma to ensure that your kernel will boot if you choose lzma compression for modules; helpful, because there's no other way for the package manager to track that dependency. There are other more appropriate places for USE=lzma: * In a profile that the user has to choose, like the desktop profile. * In the packages themselves, as IUSE defaults. * Nowhere :) All Gentoo users know how to set USE=lzma if they want lzma support. You learn this before you even boot into your first Gentoo install.