diff mbox series

[bug#40207,2/2] tests: install: Add %test-gui-installed-desktop-os-encrypted.

Message ID 87d090nwr7.fsf@gmail.com
State Accepted
Headers show
Series None | expand

Checks

Context Check Description
cbaines/applying patch fail View Laminar job

Commit Message

Mathieu Othacehe March 25, 2020, 5:57 p.m. UTC 
Hey,

> What about making a separate OS definition for the desktop installation
> case?  We’d move the ‘os’ variable to the top-level and add an ‘os’
> parameter to ‘guided-installation-test’.
>
> Having two separate OS definitions (one for desktop install, one for
> “bare-bones” install) should be clearer than having a single definition
> with conditionals.

I fixed all your remarks. But there's still an issue, the size of the
installation image is guessed badly in "run-install" for
"%test-gui-installed-desktop-os-encrypted"

--8<---------------cut here---------------start------------->8---
(image  (system-disk-image
         (operating-system-with-gc-roots
          os (list target))
         #:disk-image-size 'guess ;guess is too low.
         #:file-system-type
         installation-disk-image-file-system-type)))
--8<---------------cut here---------------end--------------->8---

applying the dirty snippet below fixes it, but you may have a better idea?

--8<---------------cut here---------------start------------->8---
Thanks,

Mathieu

Comments

Ludovic Courtès March 25, 2020, 9:41 p.m. UTC | #1 
Hi!

Mathieu Othacehe <m.othacehe@gmail.com> skribis:

>> What about making a separate OS definition for the desktop installation
>> case?  We’d move the ‘os’ variable to the top-level and add an ‘os’
>> parameter to ‘guided-installation-test’.
>>
>> Having two separate OS definitions (one for desktop install, one for
>> “bare-bones” install) should be clearer than having a single definition
>> with conditionals.
>
> I fixed all your remarks. But there's still an issue, the size of the
> installation image is guessed badly in "run-install" for
> "%test-gui-installed-desktop-os-encrypted"
>
> (image  (system-disk-image
>          (operating-system-with-gc-roots
>           os (list target))
>          #:disk-image-size 'guess ;guess is too low.
>          #:file-system-type
>          installation-disk-image-file-system-type)))
>
>
> applying the dirty snippet below fixes it, but you may have a better idea?
>
> --- a/gnu/build/vm.scm
> +++ b/gnu/build/vm.scm
> @@ -233,7 +233,7 @@ deduplicates files common to CLOSURE and the rest of PREFIX."
>    "Return the estimated size of a partition that can store the store items
>  given by GRAPHS, a list of file names produced by #:references-graphs."
>    ;; Simply add a 25% overhead.
> -  (round (* 1.25 (closure-size graphs))))
> +  (round (* 1.30 (closure-size graphs))))

Weird, the 25% hack had always worked fine.

I don’t have a better idea though.  Perhaps it’s better to pass an
explicit size in ‘run-install’ than to modify the 25% hack above?

Thanks,
Ludo’.
Mathieu Othacehe March 26, 2020, 10:54 a.m. UTC | #2 
Hey,

> I don’t have a better idea though.  Perhaps it’s better to pass an
> explicit size in ‘run-install’ than to modify the 25% hack above?

Seems fair, fixed and pushed!

Thanks,

Mathieu
diff mbox series

Patch

--- a/gnu/build/vm.scm
+++ b/gnu/build/vm.scm
@@ -233,7 +233,7 @@  deduplicates files common to CLOSURE and the rest of PREFIX."
   "Return the estimated size of a partition that can store the store items
 given by GRAPHS, a list of file names produced by #:references-graphs."
   ;; Simply add a 25% overhead.
-  (round (* 1.25 (closure-size graphs))))
+  (round (* 1.30 (closure-size graphs))))

 (define* (initialize-partition-table device partitions
--8<---------------cut here---------------end--------------->8---