Bug 873541 – app-emacs/flycheck: ~alpha keywording request

• Status: RESOLVED FIXED
• Product: Gentoo Linux

[Description matoro 2022-09-30 00:13:53 UTC]

Thanks

[Comment 1 Sam James 2022-09-30 00:16:38 UTC]

USE=emacs is an example of something we should just use.mask for now.

[Comment 2 matoro 2022-09-30 00:22:31 UTC]

(In reply to Sam James from comment #1)
> USE=emacs is an example of something we should just use.mask for now.

Sorry, why is that?  Does emacs not work or something?  I wouldn't know what to put in the note.

[Comment 3 Ulrich Müller 2022-09-30 07:39:21 UTC]

(In reply to Sam James from comment #1)
> USE=emacs is an example of something we should just use.mask for now.

Why is this? Are any of the packages in the list not labelled allarches? (I haven't checked, but I would be surprised.)

[Comment 4 Michał Górny 2022-09-30 07:42:32 UTC]

(In reply to Ulrich Müller from comment #3)
> (In reply to Sam James from comment #1)
> > USE=emacs is an example of something we should just use.mask for now.
> 
> Why is this? Are any of the packages in the list not labelled allarches? (I
> haven't checked, but I would be surprised.)

ALLARCHES are for subsequent stabilizations, not first keywording.

[Comment 5 Ulrich Müller 2022-09-30 09:02:19 UTC]

(In reply to Michał Górny from comment #4)
> ALLARCHES are for subsequent stabilizations, not first keywording.

Which I never understood. Either a package is arch independent or it isn't.

[Comment 6 Sam James 2022-09-30 22:24:56 UTC]

(In reply to Ulrich Müller from comment #3)
> (In reply to Sam James from comment #1)
> > USE=emacs is an example of something we should just use.mask for now.
> 
> Why is this? Are any of the packages in the list not labelled allarches? (I
> haven't checked, but I would be surprised.)

I explained this a bit w/ matoro on IRC, the idea being that exp profile status are usually b/c of lack of resources and keywording something tends to have a knock-on effect and requires sustained work for new deps and so on.

USE=emacs in that view is, for the time being, a luxury and not really worth bothering with until the depgraph is consistent and until we can assess what is/isn't worth spending the time on. Especially given there's no official Gentoo alpha HW right now.

[Comment 7 Ulrich Müller 2022-10-01 06:33:51 UTC]

(In reply to Sam James from comment #6)
> USE=emacs in that view is, for the time being, a luxury and not really worth
> bothering with until the depgraph is consistent and until we can assess what
> is/isn't worth spending the time on. Especially given there's no official
> Gentoo alpha HW right now.

Sorry, but that doesn't make any sense. These packages are truly arch-independent and will require zero effort after initial keywording. Use-masking emacs would cripple the editor's functionality (because of missing support packages) for no good reason. It would also cause inconsistent behaviour on different archs which is the last thing we want for app-emacs.

[Comment 8 Sam James 2022-10-01 06:35:31 UTC]

(In reply to Ulrich Müller from comment #7)
> (In reply to Sam James from comment #6)
> > USE=emacs in that view is, for the time being, a luxury and not really worth
> > bothering with until the depgraph is consistent and until we can assess what
> > is/isn't worth spending the time on. Especially given there's no official
> > Gentoo alpha HW right now.
> 
> Sorry, but that doesn't make any sense. These packages are truly
> arch-independent and will require zero effort after initial keywording.
> Use-masking emacs would cripple the editor's functionality (because of
> missing support packages) for no good reason. It would also cause
> inconsistent behaviour on different archs which is the last thing we want
> for app-emacs.

But it does make sense given the above mechanism for which we use keywording -- i.e. we don't propagate first-time keywording via ALLARCHES.

If you change the procedure, then yes, we can change how we treat it.

[Comment 9 Michał Górny 2022-10-01 06:42:03 UTC]

(In reply to Ulrich Müller from comment #7)
> Sorry, but that doesn't make any sense. These packages are truly
> arch-independent and will require zero effort after initial keywording.
> Use-masking emacs would cripple the editor's functionality (because of
> missing support packages) for no good reason. It would also cause
> inconsistent behaviour on different archs which is the last thing we want
> for app-emacs.

So none of them use emacs-lisp then?  Because as soon as they're not plain data, and data in a format that's really well-tested on given arch you're opening to unfound bugs in compilers, interpreters, etc.  The purpose of testing is to actually test stuff, not rubber-stump because "it should work".

[Comment 10 Ulrich Müller 2022-10-01 07:19:17 UTC]

(In reply to Sam James from comment #8)
> But it does make sense given the above mechanism for which we use keywording
> -- i.e. we don't propagate first-time keywording via ALLARCHES.
> 
> If you change the procedure, then yes, we can change how we treat it.

We don't use.mask emacs on any other of the dev archs. For the exp archs (where app-editors/emacs itself is keyworded), we do mask it on m68k but not on mips.

So, if the alpha team doesn't even have enough resources to do a simple initial keywording of a couple of packages that have very small compile times, then maybe alpha should stay exp for the time being?

[Comment 11 Sam James 2022-10-01 07:21:31 UTC]

(In reply to Ulrich Müller from comment #10)
> (In reply to Sam James from comment #8)
> > But it does make sense given the above mechanism for which we use keywording
> > -- i.e. we don't propagate first-time keywording via ALLARCHES.
> > 
> > If you change the procedure, then yes, we can change how we treat it.
> 
> We don't use.mask emacs on any other of the dev archs. For the exp archs
> (where app-editors/emacs itself is keyworded), we do mask it on m68k but not
> on mips.
> 
> So, if the alpha team doesn't even have enough resources to do a simple
> initial keywording of a couple of packages that have very small compile
> times, then maybe alpha should stay exp for the time being?

The depgraph is going to continue to grow as USE=emacs has to be keyworded for various packages. We don't consider Emacs obligatory for a profile to be non-exp.

But I'll defer to the arch team lead on what he wants. What I've done when moving from exp->dev is mask as much as possible, get the depgraph consistent, then review masks. Others may have different strategies.

Note that none of this changes that the keywording policy works a way you disagree with right now. If you want to change that, please do propose it.

[Comment 12 Ulrich Müller 2022-10-01 07:59:01 UTC]

(In reply to Sam James from comment #11)
> Note that none of this changes that the keywording policy works a way you
> disagree with right now. If you want to change that, please do propose it.

A special "noarch" keyword was discussed several times in the past but went nowhere (and I tend to agree that such a thing isn't feasible). So ALLARCHES was established as a substitute, but we went only half the way there because we didn't extend it to keywording.

What I would really love is to keyword all app-emacs packages (with a few exceptions that e.g. contain C parts) for all keywords of Emacs itself, and add new packages with a complete set of keywords. This would spare us other discussions like this in the future.


(In reply to Michał Górny from comment #9)
> So none of them use emacs-lisp then?  Because as soon as they're not plain
> data, and data in a format that's really well-tested on given arch you're
> opening to unfound bugs in compilers, interpreters, etc.  The purpose of
> testing is to actually test stuff, not rubber-stump because "it should work".

By that argument, should we also drop keywords when doing a major version bump of a package? For example, I carried ~m68k over from app-editors/emacs:27 to :28. I feel much more uneasy about that one than about allarches-keywording of any app-emacs package.

[Comment 13 Michał Górny 2022-10-01 08:06:03 UTC]

(In reply to Ulrich Müller from comment #12)
> By that argument, should we also drop keywords when doing a major version
> bump of a package? For example, I carried ~m68k over from
> app-editors/emacs:27 to :28. I feel much more uneasy about that one than
> about allarches-keywording of any app-emacs package.

Of course.  If you feel like there's a major breakage potential, you should start with new keyword set.  Isn't this in ebuild quiz, btw?

[Comment 14 Ulrich Müller 2022-10-01 08:50:53 UTC]

(In reply to Michał Górny from comment #13)
> Of course.  If you feel like there's a major breakage potential, you should
> start with new keyword set.  Isn't this in ebuild quiz, btw?

It's in the ebuild quiz, which refers to the devmanual, where I cannot find such a statement. But maybe that's just me. :/

[Comment 15 Larry the Git Cow 2022-10-06 02:24:47 UTC]

The bug has been closed via the following commit(s):

https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=a8e2241d807ae940c3d05b75ca98d6e6b5a1502b

commit a8e2241d807ae940c3d05b75ca98d6e6b5a1502b
Author:     Matoro Mahri <[e-mail]>
AuthorDate: 2022-09-30 14:45:13 +0000
Commit:     Matt Turner <[e-mail]>
CommitDate: 2022-10-06 02:24:39 +0000

    profiles/arch/alpha: mask dev-util/pkgcheck[emacs]
    
    Per Sam, avoid pulling in app-emacs/flycheck deptree.
    
    Closes: https://bugs.gentoo.org/873541
    Closes: https://github.com/gentoo/gentoo/pull/27539
    Signed-off-by: Matoro Mahri <[e-mail]>
    Signed-off-by: Matt Turner <[e-mail]>

 profiles/arch/alpha/package.use.mask | 4 ++++
 1 file changed, 4 insertions(+)

[Comment 16 matoro 2022-10-06 04:07:52 UTC]

Reopened since this discussion is still ongoing.  Leaving the package.use.mask in, will work on keywording these anyway.

[Comment 17 matoro 2022-10-08 15:59:12 UTC]

alpha done

all arches done

[Comment 18 Larry the Git Cow 2022-11-10 16:55:29 UTC]

The bug has been closed via the following commit(s):

https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=67c9871d875f6a97b44e053b5d24ef27c71e446c

commit 67c9871d875f6a97b44e053b5d24ef27c71e446c
Author:     Matoro Mahri <[e-mail]>
AuthorDate: 2022-10-08 18:11:44 +0000
Commit:     Mike Gilbert <[e-mail]>
CommitDate: 2022-11-10 16:55:17 +0000

    profiles/arch/alpha: unmask dev-util/pkgcheck[emacs]
    
    Closes: https://bugs.gentoo.org/873541
    Signed-off-by: Matoro Mahri <[e-mail]>
    Closes: https://github.com/gentoo/gentoo/pull/27698
    Signed-off-by: Mike Gilbert <[e-mail]>

 profiles/arch/alpha/package.use.mask | 4 ----
 1 file changed, 4 deletions(-)