Ludovic Courtès writes:

Hi!

> "Jan (janneke) Nieuwenhuizen" <janneke@gnu.org> skribis:
>
>> * gnu/build/hurd-boot.scm (make-hurd-device-nodes): Do not create
>> dev/{null,zero,full,random,urandom} mount points.
>> (setup-translator, xattr-translator,
>> showtrans-translator?, translated?, set-hurd-device-translators): New
>> procedures.
>> (boot-hurd-system): Use them instead of running MAKEDEV.
>
> This is mostly about moving the logic from the MAKEDEV script to this
> file, right?  Sounds nice.

Yes.  Glad you like the idea too.

>> +(define (xattr-translator? file-name)
>> +  "Return true if FILE-NAME has an extended @code{gnu.translator} attribute
>> +set."
>> +  (false-if-exception
>> +   (not (string-null? (getxattr file-name "gnu.translator")))))
>
> I’d call it ‘passive-translator-xattr?’.

> In general, I’d avoid ‘false-if-exception’ as much as possible because
> it can hide real issues.  So here, you could catch 'system-error and
> check for ENODATA.  If you need it in several places, you can define a
> ‘false-if-ENODATA’ macro

Nice.  Ah, I remember having file-exists? checks or no check at all and
then found it only worked on first boot, not with a persistent image.
Minor detail.  So I got out my axe and..."it worked" :-/

>> +(define (showtrans-translator? file-name)
>> +  "Return true if @file{showtrans} finds a translator installed on FILE-NAME."
>
> Should be ‘passive-translator-installed?’ no?  (IIRC ‘showtrans’ only
> shows passive translator settings by default.)

Yes, thanks.

>> +  (with-output-to-port (%make-void-port "w")
>> +    (lambda _
>> +      (with-error-to-port (%make-void-port "w")
>> +        (lambda _
>> +          (zero? (system* "showtrans" "--silent" file-name)))))))
>> +
>> +(define (translated? file-name)
>> +  "Return true if a translator is installed on FILE-NAME."
>> +  (if (string-contains %host-type "linux-gnu")
>> +      (xattr-translator? file-name)
>> +      (showtrans-translator? file-name)))
>
> It’s counter-intuitive that hurd-boot.scm is used from GNU/Linux.
> Should we move the shared bits in (gnu build hurd) or similar?

Maybe...I don't know.  hurd-boot defines functions that are needed to
create a bootable hurd -- some can be called from GNU/Linux.

Similarly, "linux-boot" defines "make-essential-device-nodes", which
could be called from GNU/Hurd when we cross build a linux VM.

Well, having one weird situation is a bad argument to propagate the
madness.

I guess it makes sense to move utility functions like
passive-translator-xattr, passive-translator-installed?, and translated?
could be moved to (gnu build hurd).

Haven't made this change yet, let me know you want; or feel free to make
the change yourself :-)

>> +(define* (setup-translator file-name command #:optional (mode #o600))
>> +  "Setup translator COMMAND on FILE-NAME."
>
> ‘set-translator’?  :-)

=> set-translator! :-)

>> +  (false-if-exception (mkdir-p (scope "dev/vcs/1")))
>> +  (false-if-exception (mkdir-p (scope "dev/vcs/2")))
>> +  (false-if-exception (mkdir-p (scope "dev/vcs/3")))
>> +  (false-if-exception (rename-file "/dev/console" "/dev/console-"))
>> +  (for-each scope-setup-translator devices)
>> +
>> +  (false-if-exception (symlink "/dev/random" "/dev/urandom"))
>> +  (false-if-exception (mkdir-p "/dev/fd"))
>> +  (false-if-exception (symlink "/dev/fd/0"   "/dev/stdin"))
>> +  (false-if-exception (symlink "/dev/fd/1"   "/dev/stdout"))
>> +  (false-if-exception (symlink "/dev/fd/2"   "/dev/stderr")))
>
> ‘false-if-EEXIST’?

Done for symlink; for mkdir, using

  (define (mkdir* dir)
    (let ((dir (scope dir)))
     (unless (file-exists? dir)
       (mkdir-p dir))))

New patch attached.

Greetings,
Janneke