Re: Use of null (begin) blocks is deprecated in Guile stable V2.0, use unspecified? and/or *unspecified* instead.

[David Kastrup]

David Kastrup <address@hidden> writes:

> Ian Hulin <address@hidden> writes:
>
>> This is forwarded from the Guile bug list.  Bug-squad please create a
>> LilyPond issue for this - we need to change our code:
>>
>> Files affected appear (according to git grep) to be:
>> lily/mensural-ligature.cc
>> lily/system.cc
>> scm/c++.scm
>> scm/define-event-classes.scm
>> scm/music-functions.scm
>
> I am responsible for most of those, and will fix them.  No issue
> required here.
>
> For my defense, neither *unspecified* nor unspecified? are documented in
> the Guile-1.8 manual though working at the prompt.  So while the 1.8
> branch is still maintained, you should report this as a documentation
> bug.  And after all, it _does_ affect upwards-compatibility.

As a meta-issue: I had defined void? as something quite equivalent to
Guile's unspecified?.  I am retaining this definition since Guile's
"unspecified?", meaning eq? to the specific value *unspecified*, is a
misleading name.  And the context we use it for, namely checking whether
an expression does not return any value, is quite specific.

define-void-function defines a function returning no value, and I would
not want to call this define-unspecified-function instead.

I consider the idea of making *unspecified* equivalent to (values)
eminently sensible.  I don't understand why this would need to be
different from a hardwired constant SCM_UNSPECIFIED internally, quite
similar to how '() can be equivalent to a hardwired constant SCM_EOL.

-- 
David Kastrup