Am 24.10.2017 um 23:34 schrieb Ludovic Courtès:
> Indeed.  I figured something like this works:
> 
> --8<---------------cut here---------------start------------->8---
> scheme@(guile-user)> (define-syntax match-record
> 		       (syntax-rules ()
> 			 ((_ record type (field fields ...) body ...)
> 			  (if (eq? (struct-vtable record) type)
> 			      (let ((field ((record-accessor type 'field) record)))
> 				(match-record record type (fields ...) body ...))
> 			      (throw 'wrong-type-arg record)))
> 			 ((_ record type () body ...)
> 			  (begin body ...))))
> scheme@(guile-user)> (match-record coreutils (@@ (guix packages) <package>) (home-page) home-page)
> $6 = "https://www.gnu.org/software/coreutils/"
> scheme@(guile-user)> (match-record coreutils (@@ (guix packages) <package>) (home-page synopsis) (list synopsis home-page))
> $7 = ("Core GNU utilities (file, text, shell)" "https://www.gnu.org/software/coreutils/")
> --8<---------------cut here---------------end--------------->8---

Great!

> 
> We could use that for now.
> 
> Eventually though, we should have something better in (guix records)
> that (1) computes indices and report wrong-field-name errors at
> expansion time, and (2) accounts for thunked/delayed fields.
> 
> WDYT?

I didn't even know guix records had those features :)

> 
> If the above macro is good enough, we can add it to (guix records) with
> a TODO comment.  That would already be better than the other options.
> 

I added it for now. Personally I don't like having functions with big
TODOs like this. What would be the solution for thunked delayed fields?
Force them as they are bound in the let?


>> I also noticed a missing equal sign after rememberchannel in the
>> defaultconfig and added that.
> 
> I noticed a couple of obvious mistakes:
> 
> 
> 
> diff --git a/gnu/services/telephony.scm b/gnu/services/telephony.scm
> index 0c30b409f..a305a1be8 100644
> --- a/gnu/services/telephony.scm
> +++ b/gnu/services/telephony.scm
> @@ -240,7 +240,7 @@ Or set public-registration to #f")))))))))
>  (define (murmur-activation config)
>    #~(begin
>        (use-modules (guix build utils))
> -      (let ((log-dir (dirname #$(murmur-configuration-log-file config)))
> +      (let* ((log-dir (dirname #$(murmur-configuration-log-file config)))
>               (pid-dir (dirname #$(murmur-configuration-pid-file config)))
>               (db-dir (dirname #$(murmur-configuration-database-file config)))
>               (user (getpwnam #$(murmur-configuration-user config)))
I think there was no mistake here the init-dir function took the user as
argument, but I changed it into the let* form and removed the argument now.

> @@ -283,7 +283,7 @@ Or set public-registration to #f")))))))))
>           (documentation "Run the murmur mumble-server.")
>           (requirement '(networking))
>           (start #~(make-forkexec-constructor
> -                   '(#$(file-append (murmur-configuration-package)
> +                   '(#$(file-append (murmur-configuration-package config)
>                                      "/bin/murmurd")
>                       "-ini"
>                       #$(or (murmur-configuration-file config)
> 
Ouch, so much about me thinking that I could just make a quick change.

> 
> This makes me think that it would be good to have a unit test.  Would
> you like to try writing one now (see the examples in gnu/tests/*.scm),
> or do you prefer to leave it for later?
I would like to write some tests, but right now I need to setup my guix
development environment on a different computer first. On my current
setup I have 15 gigabytes of free hard drive space and when I run `make
check-system` it fails with some 'no space left on device' message.

> 
> In the latter case, please test the system to make sure it actually
> works (that can be done in a VM.)
For this patch:
I ran make and got no warnings.
I deployed it on my server and connected with mumble from my computer
and it worked.