notmuch: extract optional dependencies #84008

mredaelli · 2020-04-01T17:42:36Z

Motivation for this change

I realized today that installing neomutt installs notmuch which installs emacs which to me, using neovim, seems a bit wasteful :)

I extracted the optional dependencies and put them behind a few package options. Using this as an overlay on my system works, but would be happy to discuss everything, from formatting to defaults of the optional parameters to testing/touching other derivations depending on this one, ...

Things done

puckipedia · 2020-04-01T23:09:27Z

I believe the standard way of doing optional dependencies is to make them optional in the attrs (so ? null). For with-docs i think using withDocs would be more canonical?

mredaelli · 2020-04-02T08:19:48Z

I believe the standard way of doing optional dependencies is to make them optional in the attrs (so ? null).

I can do that for ruby, but emacs for some reason, AFAIK, is needed for the tests, even if you skip the emacs tests and even if it's compiled without emacs, so we can't make it null.

Do you see a way around that?

For with-docs i think using withDocs would be more canonical?

Absolutely. Will fix that.

Mic92 · 2020-04-02T09:37:43Z

pkgs/applications/networking/mailreaders/notmuch/default.nix

@@ -7,6 +7,8 @@
 , emacs
 , ruby
 , which, dtach, openssl, bash, gdb, man
+, withDocs ? true


How about having a different output for docs instead of disabling it?

I'm afraid I don't know what that means/how to do it ^_^

You basically add:

outputs = [ "out" "doc" ];

Than it should move "$out/usr/share/doc" to "$doc" instead of "$out".
Having multiple outputs is a feature of nixpkgs and described in the nixpkgs manual here: https://nixos.org/nixpkgs/manual/#chap-multiple-output

Something like this?

Mic92 · 2020-04-02T14:05:44Z

pkgs/applications/networking/mailreaders/notmuch/default.nix

-  configureFlags = [ "--zshcompletiondir=${placeholder "out"}/share/zsh/site-functions" ];
+  configureFlags = [
+    "--zshcompletiondir=${placeholder "out"}/share/zsh/site-functions"
+    "--mandir=${placeholder "doc"}"


Maybe this should go to man output. Some people want to read the manpage without reading the developer documentation. Also the man output automatically picks up the right files without an need for having to specify its path in configureFlags.

yeah I noticed right after pushing. Fixing it now

flokli · 2020-04-02T20:11:17Z

pkgs/applications/networking/mailreaders/notmuch/default.nix

-    bash-completion  # (optional) dependency to install bash completion
-    emacs  # (optional) to byte compile emacs code, also needed for tests
-    ruby  # (optional) ruby bindings
+    bash-completion           # (optional) dependency to install bash completion


Is this invoked during build, or made a runtime reference? If the former is the case, it'd need to be in nativeBuildInputs.

Hmm likely build only. Let me try

Seems to work. I updated the PR

Meh, I quickly tried by cross-compiling neomutt to aarch64, but this failed, as talloc required for notmuch currently fails to cross-compile.

I assume this is fine, we can always fix it up later.

flokli · 2020-04-03T15:58:06Z

Thanks!

xaverdh · 2020-04-04T07:18:35Z

Just out of curiosity: In the original derivation, does emacs end up in the runtime closure just due to the substituteInPlace part? Then I guess this could be solved by putting notmuch-emacs-mua into a different output (assuming that nix tracks runtime dependencies per output)?

mredaelli · 2020-04-04T17:14:44Z

AFAIK, it happened because emacs was in the buildInputs

flokli · 2020-04-04T18:22:23Z

Nix tracks runtime closures, and does that per output. Even if you'd add libreoffice to buildInputs, if there's no reference in the "binary output", it wouldn't be downloaded if you'd donwload it from the cache (it however would be downloaded if you want to build it by yourself).

For this specific package, there's a single reference to emacs in the $out output:

nix why-depends -a /nix/store/d2zz5rkh9rdf6fqjbpx1mfdzzl6n7ghj-notmuch-0.29.3 /nix/store/1rx6bbd0fcw7h0shz1drrhza1flpvh0w-emacs-26.3
/nix/store/d2zz5rkh9rdf6fqjbpx1mfdzzl6n7ghj-notmuch-0.29.3
╚═══bin/notmuch-emacs-mua: …}".}..EMACS=${EMACS:-/nix/store/1rx6bbd0fcw7h0shz1drrhza1flpvh0w-emacs-26.3/bin/emacs}.EMACSCLIE…
    => /nix/store/1rx6bbd0fcw7h0shz1drrhza1flpvh0w-emacs-26.3

if $out/bin/notmuch-emacs-mua would be moved to another output, it wouldn't be downloaded if you would only download the out output from the binary cache.

I'm not exactly sure on how emacs integrates with notmuch works, and if moving bin/notmuch-emacs-mua would break emacs. spawning @tazjin and @andir on that ;-)

tazjin · 2020-04-04T18:29:52Z

I'm not exactly sure on how emacs integrates with notmuch works, and if moving bin/notmuch-emacs-mua would break emacs. spawning @tazjin and @andir on that ;-)

notmuch-emacs-mua is used as the handler for mailto: links when other applications want to spawn the user's mail client. This will need to be propagated to the user's environment if they're using the standard notmuch client (i.e. notmuch.el).

flokli · 2020-04-04T18:52:03Z

`notmuch-emacs-mua` is used as the handler for `mailto:` links when other applications want to spawn the user's mail client. This will need to be propagated to the user's environment if they're using the standard notmuch client (i.e. notmuch.el).

Thanks for the explaination! So if we'd move `notmuch-emacs-mua` to a separate `emacs` output and add some release notes explaining it now needs to be obtained from the `notmuch.emacs` output, this should be fine.

tazjin · 2020-04-04T19:42:39Z

So if we'd move notmuch-emacs-mua to a separate emacs output and add some release notes explaining it now needs to be obtained from the notmuch.emacs output, this should be fine.

If you're splitting it, then the Emacs client itself should also be moved to that output to avoid confusion.

flokli · 2020-04-04T21:54:05Z

On 20-04-04 12:42:52, Vincent Ambo wrote: > So if we'd move `notmuch-emacs-mua` to a separate `emacs` output and add some release notes explaining it now needs to be obtained from the `notmuch.emacs` output, this should be fine. If you're splitting it, then the Emacs client itself should also be moved to that output to avoid confusion.

You mean `$out/share/emacs`, right? I don't see any other emacs-related things in `$out`…

andir · 2020-04-06T09:51:25Z

On 14:54 04.04.20, Florian Klink wrote: On 20-04-04 12:42:52, Vincent Ambo wrote: >> So if we'd move `notmuch-emacs-mua` to a separate `emacs` output and add some release notes explaining it now needs to be obtained from the `notmuch.emacs` output, this should be fine. > >If you're splitting it, then the Emacs client itself should also be moved to that output to avoid confusion. You mean `$out/share/emacs`, right? I don't see any other emacs-related things in `$out`…

Just my 2 cents. Haven't actually opend the GH page for this PR so I only got what GH decided to send via mail. I don't know an example from the top of my head but I think for other plugin systems we have code that basically does `plugin.someEditor or plugin` to automatically convert a package that is a plugin for `someEditor` (or `someSoftware`) to a plugin (or the path of the plugin). Don't we have a similar system for emacs? If not, wouldn't that make sense? I sounds better then having to ask each user update their configuraiton just because some binary build also supports a plugin for some editor. Also it should keep the domain knowledge (of where the plugin code in the notmuch package is) to the notmuch package.

tazjin · 2020-04-06T10:17:15Z

Don't we have a similar system for emacs?

Seems like something @adisbladis would know.

flokli · 2020-04-07T20:08:31Z

There seems to be emacsPackages.notmuch, but these seem to come from another source, and don't seem to have a connection to pkgs.notmuch.

ofborg bot requested review from flokli, the-kenny and puckipedia April 1, 2020 17:48

ofborg bot added 10.rebuild-darwin: 11-100 10.rebuild-linux: 11-100 labels Apr 1, 2020

mredaelli force-pushed the notmuch branch from 10d3a96 to d215074 Compare April 2, 2020 08:58

Mic92 reviewed Apr 2, 2020

View reviewed changes

mredaelli force-pushed the notmuch branch from d215074 to 791bf75 Compare April 2, 2020 14:00

Mic92 reviewed Apr 2, 2020

View reviewed changes

mredaelli force-pushed the notmuch branch from 791bf75 to a42ec01 Compare April 2, 2020 14:11

flokli reviewed Apr 2, 2020

View reviewed changes

notmuch: extract optional dependencies

b6f6fd8

mredaelli force-pushed the notmuch branch from a42ec01 to b6f6fd8 Compare April 3, 2020 14:41

ofborg bot requested a review from flokli April 3, 2020 14:48

flokli merged commit ed3ec0b into NixOS:master Apr 3, 2020

flokli mentioned this pull request Apr 7, 2020

notmuch: set emacs off by default #84610

Closed

10 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

notmuch: extract optional dependencies #84008

notmuch: extract optional dependencies #84008

mredaelli commented Apr 1, 2020 •

edited

puckipedia commented Apr 1, 2020

mredaelli commented Apr 2, 2020

Mic92 Apr 2, 2020

mredaelli Apr 2, 2020

Mic92 Apr 2, 2020

mredaelli Apr 2, 2020

Mic92 Apr 2, 2020

mredaelli Apr 2, 2020

mredaelli Apr 2, 2020

flokli Apr 2, 2020

mredaelli Apr 3, 2020

mredaelli Apr 3, 2020

flokli Apr 3, 2020

flokli commented Apr 3, 2020

xaverdh commented Apr 4, 2020

mredaelli commented Apr 4, 2020

flokli commented Apr 4, 2020

tazjin commented Apr 4, 2020

flokli commented Apr 4, 2020 via email

tazjin commented Apr 4, 2020

flokli commented Apr 4, 2020 via email

andir commented Apr 6, 2020 via email

tazjin commented Apr 6, 2020

flokli commented Apr 7, 2020

notmuch: extract optional dependencies #84008

notmuch: extract optional dependencies #84008

Conversation

mredaelli commented Apr 1, 2020 • edited

Motivation for this change

Things done

puckipedia commented Apr 1, 2020

mredaelli commented Apr 2, 2020

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

flokli commented Apr 3, 2020

xaverdh commented Apr 4, 2020

mredaelli commented Apr 4, 2020

flokli commented Apr 4, 2020

tazjin commented Apr 4, 2020

flokli commented Apr 4, 2020 via email

tazjin commented Apr 4, 2020

flokli commented Apr 4, 2020 via email

andir commented Apr 6, 2020 via email

tazjin commented Apr 6, 2020

flokli commented Apr 7, 2020

mredaelli commented Apr 1, 2020 •

edited