Re: What's up with 'current-load-port'

[Thomas Morley <thomasmorley65@gmail.com>]

Am Mo., 5. Nov. 2018 um 02:43 Uhr schrieb Mark H Weaver <mhw@netris.org>:
>
> Thomas Morley <thomasmorley65@gmail.com> writes:
>
> > what's up with 'current-load-port'?
> >
> > Simply checking in a guile-prompt I get:
> > guile-1.8: #<primitive-procedure current-load-port>
> > guile-2.0.14: #<procedure current-load-port ()>
> > guile-2.2.4 and guile-2.9.1:
> > ;;; <unknown-location>: warning: possibly unbound variable `current-load-port'
> > ERROR: In procedure module-lookup: Unbound variable: current-load-port
> >
> > It's in the manual, though, without any hint it could be deprecated or
> > disabled or the need to use a certain module.
>
> Good catch.  I agree that this was mishandled.
>
> Out of curiosity, what do you use 'current-load-port' for?
>
> Although it's in the manual, the description doesn't really describe
> what the procedure returns, or when one can rely upon it being set.  It
> only says that it's "used internally by 'primitive-load'".
>
>       Mark

Some background:
You probably remember I'm from LilyPond. Currently we still use
guilev-1.8, though have experimental support for guilev2.
I do a lot of support on our user-list, frequently checking things
with my lilypond-guilev2-setup to detect bugs.

Recently a user wrote a large scm-file and put it into lilypond using
'(load file.scm)'.
No problem with guilev1, but with guilev2 'file.scm' is not found.
Thus I tried to track the problem down.
Not a bad start is to rtfm :)
So I found 'current-load-port' and intended to play around with it
whether I can get some useful info out of it.

In short, I likely don't need 'current-load-port' for anything, but
couldn't be sure before I used it. Thus this thread and patch.

While doing rtfm I found 'primitive-load' working for the above
described purpose in guilev1 and guilev2. So the problem is solvable.

Although this doesn't explain why 'load' stopped working for lilypond.
I then started from our master (guilev1) and changed only two things:
- teaching configure to accept guile-2.9.1
- disabling a certain not longer supported function (only needed for
collecting some statistic data and never used afair)
Omitting all other guilev2-patches will ofcourse result in a very
buggy lilypond, if someone would intend to use it, but enough to check
whether 'load' works.
It does not.
So the underlying problem is either already in lilypond-master or
guilev2 changed something in 'load'.

That's my current research-state.

Any hint from the guile side?


Thanks,
  Harm