[bug#74860] [PATCH 02/10] services: Add ‘log-rotation-service-type’.

[Maxim Cournoyer <maxim.cournoyer@gmail.com>]

Hi Ludovic,

Ludovic Courtès <ludo@gnu.org> writes:

> * gnu/services/admin.scm (%default-log-rotation-calendar-event): New
> variable.
> (<log-rotation-configuration>): New record type.
> (log-rotation-shepherd-services): New procedure.
> (log-rotation-service-type): New variable.

Neat.

[...]

> +@item @code{compression} (default: @code{'gzip})
> +This is the compression method used for rotated log files, one of
> +@code{'none}, @code{'gzip}, and @code{'zstd}.

Why not default to zstd?  It's superior to gzip in all regards, as far
as I know.  We use it already preferably to gzip to compress man pages
for example.  Perhaps we should consider it as our default compressor
from now on?

[...]

> +\f
> +;;;
> +;;; Shepherd's log rotation service.
> +;;;
> +
> +(define %default-log-rotation-calendar-event
> +  ;; Default calendar event when log rotation is triggered.
> +  #~(calendar-event #:minutes '(0)
> +                    #:hours '(22)
> +                    #:days-of-week '(sunday)))
> +
> +(define-record-type* <log-rotation-configuration>
> +  log-rotation-configuration make-log-rotation-configuration
> +  log-rotation-configuration?
> +  (provision      log-rotation-configuration-provision (default '(log-rotation)))
> +  (requirement    log-rotation-configuration-requirement
> +                  (default (if for-home? '() '(user-processes))))
> +  (calendar-event log-rotation-configuration-calendar-event
> +                  (default %default-log-rotation-calendar-event))
> +  (external-log-files log-rotation-configuration-external-log-files
> +                      (default '()))
> +  (compression    log-rotation-configuration-compression (default 'gzip))
> +  (expiry         log-rotation-configuration-expiry
> +                  (default #~(%default-log-expiry)))
> +  (size-threshold log-rotation-configuration-size-threshold
> +                  (default #~(%default-rotation-size-threshold))))

Is there a technical reason for not using define-configuration here?
Otherwise, define-configuration offers superior error checking and
useful error messages on invalid input types.

If possible, I think it'd be nice to see this done with
define-configuration, for the reasons I explained above.  Otherwise it
LGTM.  Could you please send a v2 addressing this?

-- 
Thanks,
Maxim