Don’t attempt to create a module database when none are available

The present code breaks with a non-modular kernel. This can be easily avoided.

--- a/gnu/system.scm

+++ #<buffer system.scm>

@@ -477,13 +477,19 @@

value of the SYSTEM-SERVICE-TYPE service."

(let ((locale (operating-system-locale-directory os)))

(mlet* %store-monad ((kernel -> (operating-system-kernel os))

+ (kernel-modules (package-file kernel "lib/modules"))

(modules ->

(operating-system-kernel-loadable-modules os))

+ (has-modules ->

+ (or (not (null? modules))

+ (file-exists? kernel-modules)))

(kernel

(profile-derivation

(packages->manifest

(cons kernel modules))

- #:hooks (list linux-module-database)))

+ #:hooks (if has-modules

+ (list linux-module-database)

+ '())))

(initrd -> (operating-system-initrd-file os))

(params (operating-system-boot-parameters-file os)))

(return `(("kernel" ,kernel)

Thanks!

Pushed to guix master as commit b2fff3b5de7d510fe4809e9a97089dddf2a39ffc.

Hi,

Ivan Kozlov <kanichos@yandex.ru> skribis:

There were a couple of mistakes in this patch (which is fine, we all

make mistakes!), so I reverted it along with explanations:

https://git.savannah.gnu.org/cgit/guix.git/commit/?id=71977a6c59460e029c91741c8b11f3a544a70486

What problem were you trying to solve, Ivan? What do you mean by

“non-modular kernel”?

Thanks,

Ludo’.

tags 40430 + moreinfo

quit

The intent was to allow a Linux package with CONFIG_MODULES=n to work after 5c79f238634c5adb6657f1b4b1bb4ddb8bb73ef1. There are no modules and no ‘lib/modules’.

Will you accept a patch that changes linux-module-database to skip unsuitable inputs instead?

21.04.2020, 01:12, "Ludovic Courtès" <ludo@gnu.org>: