| Message ID | 20200606091552.32625-1-othacehe@gnu.org |
|---|---|
| State | Accepted |
| Headers | show |
| Series | Remove X libraries from bare-bones system closure. | expand |
| Context | Check | Description |
|---|---|---|
| cbaines/comparison | success | View comparision |
| cbaines/git branch | success | View Git branch |
| cbaines/applying patch | success | View Laminar job |
| cbaines/comparison | success | View comparision |
| cbaines/git branch | success | View Git branch |
| cbaines/applying patch | success | View Laminar job |
Mathieu Othacehe writes: Hello! > * gnu/packages/ssh.scm (openssh-no-x): New variable. [..] > diff --git a/gnu/packages/ssh.scm b/gnu/packages/ssh.scm [..] > +;; OpenSSH without X support. This allows to use OpenSSH without dragging X > +;; libraries to the closure. > +(define-public openssh-no-x > + (package > + (inherit openssh) > + (name "openssh-no-x") > + (inputs (alist-delete "xauth" (package-inputs openssh))))) I like it! Maybe use -sans-x? Two remarks: after merging, we may want to use openssh-no-x for the Hurd too and drop this from openssh proper: ,@(if (hurd-target?) '() `(("xauth" ,xauth))))) ; for 'ssh -X' and 'ssh -Y' We have this bit because xauth does not compile for the Hurd yet, but it's kind of a kludge. Not to hold this up and maybe we should take this discussion elsewhere...but I have a ~5y old, bit-rotting stack of LilyPond cross build packages (foo-SANS-X) laying around. At the time--much to my suprise--I found the LilyPond community less than enthousiastic to move their binary package builds to Guix (yet); but adding these -NO-X packages felt a bit weird. So, I wonder if we could address the X'yness of packages in a more structured way, that would be nice... Greetings, Janneke
Mathieu Othacehe <m.othacehe@gmail.com> writes: > * gnu/packages/ssh.scm (openssh-no-x): New variable. > --- > gnu/packages/ssh.scm | 8 ++++++++ > 1 file changed, 8 insertions(+) > > diff --git a/gnu/packages/ssh.scm b/gnu/packages/ssh.scm > index 73841db6b5..13cf305695 100644 > --- a/gnu/packages/ssh.scm > +++ b/gnu/packages/ssh.scm > @@ -239,6 +239,14 @@ Additionally, various channel-specific options can be negotiated.") > "See LICENSE in the distribution.")) > (home-page "https://www.openssh.com/"))) > > +;; OpenSSH without X support. This allows to use OpenSSH without dragging X > +;; libraries to the closure. > +(define-public openssh-no-x > + (package > + (inherit openssh) > + (name "openssh-no-x") > + (inputs (alist-delete "xauth" (package-inputs openssh))))) Maybe add a synopsis so users don't have to guess what this package does. I.e. (synopsis "OpenSSH client and server without X11 support"). I also think 'openssh-sans-x' is better, and is already a fairly common convention. :-)
Hello Marius, > Maybe add a synopsis so users don't have to guess what this package > does. I.e. (synopsis "OpenSSH client and server without X11 support"). > > I also think 'openssh-sans-x' is better, and is already a fairly common > convention. :-) Fixed and pushed, Thanks, Mathieu
Hey Jan, > Two remarks: after merging, we may want to use openssh-no-x for the Hurd > too and drop this from openssh proper: > > ,@(if (hurd-target?) > '() > `(("xauth" ,xauth))))) ; for 'ssh -X' and 'ssh -Y' > > We have this bit because xauth does not compile for the Hurd yet, > but it's kind of a kludge. I just did push the serie, so I think you could maybe adapt the openssh package used by the Hurd and remove the snippet above, if it's ok for you? > Not to hold this up and maybe we should take this discussion > elsewhere...but I have a ~5y old, bit-rotting stack of LilyPond cross > build packages (foo-SANS-X) laying around. At the time--much to my > suprise--I found the LilyPond community less than enthousiastic to move > their binary package builds to Guix (yet); but adding these -NO-X > packages felt a bit weird. So, I wonder if we could address the X'yness > of packages in a more structured way, that would be nice... Yes I share your concerns on that one, sadly I don't have much better to propose, and I think it's important to have a bare-bones template as minimal as possible. Anyway, this is not a big hack and it will be easy to revert it if we find a better way. Thanks, Mathieu
Mathieu Othacehe writes: Hi Mathieu, > I just did push the serie, so I think you could maybe adapt the openssh > package used by the Hurd and remove the snippet above, if it's ok for > you? Yes, sure; I'm right on it. I just sent my final patch set for wip-hurd-vm, still using plain "openssh" -- will create an and test updated patch. >> Not to hold this up and maybe we should take this discussion >> elsewhere...but I have a ~5y old, bit-rotting stack of LilyPond cross >> build packages (foo-SANS-X) laying around. At the time--much to my >> suprise--I found the LilyPond community less than enthousiastic to move >> their binary package builds to Guix (yet); but adding these -NO-X >> packages felt a bit weird. So, I wonder if we could address the X'yness >> of packages in a more structured way, that would be nice... > > Yes I share your concerns on that one, sadly I don't have much better to > propose, and I think it's important to have a bare-bones template as > minimal as possible. Anyway, this is not a big hack and it will be easy > to revert it if we find a better way. Okay...I'm sure this will come around another time. Janneke
diff --git a/gnu/packages/ssh.scm b/gnu/packages/ssh.scm index 73841db6b5..13cf305695 100644 --- a/gnu/packages/ssh.scm +++ b/gnu/packages/ssh.scm @@ -239,6 +239,14 @@ Additionally, various channel-specific options can be negotiated.") "See LICENSE in the distribution.")) (home-page "https://www.openssh.com/"))) +;; OpenSSH without X support. This allows to use OpenSSH without dragging X +;; libraries to the closure. +(define-public openssh-no-x + (package + (inherit openssh) + (name "openssh-no-x") + (inputs (alist-delete "xauth" (package-inputs openssh))))) + (define-public guile-ssh (package (name "guile-ssh")