FONTSAMPLE add option for which charset to display

[MattHeffron]

No description provided.

[nbriggs]

See comment attached.

Do you know about lispusers>fontsampler (which already had character sets as an option)?

[nbriggs]

It makes me uncomfortable to introduce a dependency on "LOADUPSDIRECTORIES" - if it really needs EXPORTS.ALL (does it?) then just (FILES EXPORTS.ALL) and it should use DIRECTORIES which should be set up in the site init file.

[MattHeffron]

So, in general, how should dependency on system level definitions (e.g., CHARSETINFO, FONTDESCRIPTOR, ...) be set in the COMS for modules in LIBRARY or LISPUSERS? The pattern:

(DECLARE%: EVAL@COMPILE DONTCOPY (FILES (FROM LOADUPS)
                                                EXPORTS.ALL))

I found in UNICODECOMS. (@rmkaplan)

[masinter]

I think the convention in the past was that those working on the "low level" system or calls to it would load the file SYSEDIT; SYSEDIT sets some variables as well as loading EXPORTS.ALL if needed. (There was also a file called "ABC" which looked like SYSEDIT and I tried to merge the two.

[MattHeffron]

See also related discussion in PR 2003

[MattHeffron]

@masinter

I think the convention in the past was that those working on the "low level" system or calls to it would load the file SYSEDIT

But ... SYSEDIT does:

(RESETVARS ((CROSSCOMPILING T))
           (FILESLOAD (SOURCE FROM LOADUPS)
                  EXPORTS.ALL))

So why wouldn't it be acceptable for individual modules (e.g., on lispusers) to do a similar FILESLOAD under (DECLARE%: EVAL@COMPILE DONTCOPY. (Without the (RESETVARS ((CROSSCOMPILING T)) which is SYSEDIT-specific.) This also wouldn't change other system-level VARS that a particular user might not want changed.

More discussion? @rmkaplan @nbriggs

[nbriggs]

@MattHeffron it only does it exactly like that because Larry changed it to do the FILESLOAD "FROM LOADUPS". It's the "FROM LOADUPS" that I'm objecting to, rather than having the right things on DIRECTORIES and not the fact of loading exports.all.

[MattHeffron]

But ... where did LOADUPSDIRECTORIES come from?
I take it that you're proposing that it be removed from the standard loadup sysouts?

[nbriggs]

It (LOADUPDIRECTORIES) is set in sources/MEDLEYDIR -- and while I'm happy to have the mechanism there (though I think MEDLEYDIR should be in library not sources), I don't believe that that is the right place to set up a lot of values that are not actually a fundamental part of the system - that's the role of the site init (site greet) file.

[MattHeffron]

So, instead of a LOADUPDIRECTORIES variable, would it be better (acceptable) for individual modules (e.g., on lispusers) to have:

(DECLARE%: EVAL@COMPILE DONTCOPY
           (FILES (SOURCE FROM VALUEOF (MEDLEYDIR "loadups>")) EXPORTS.ALL))

Should the location of EXPORTS.ALL be a known place?
The loadup sysout files are in the loadups directory because the MEDLEYDIR/scripts/medley/medley.command expects to find them in loadups.
Why not have the same a priori condition that EXPORTS.ALL is in the same place?

Or, should modules like these have a MODULENAME-SETUP-TO-COMPILE function that does the setup, or prints the setup instructions?

And, is it generally better for modules to load EXPORTS.ALL (somehow), or should they use LOADCOMP of the specific system module(s) needed? E.g., TEDIT has:

(DECLARE%: EVAL@COMPILE DONTCOPY (FILES (LOADCOMP) UNICODE))

I'm not trying to be difficult!

Should this discussion be extracted into a GitHub Issue or Discussion?

[nbriggs]

I believe it should just have (DECLARE: EVAL@COMPILE DONTCOPY (FILES EXPORTS.ALL)) - MEDLEYDIR and anything like that has anything to do with running on the emulator, unless the module is specifically targeted to something Unixy, should not appear in the module. The locations for everything should be set up in the site greet file and appear on DIRECTORIES, which FILESLOAD and friends already know about. I think that an EXPORTS.ALL that corresponds to the files that are in sources and library should be checked in to library, or at least it should appear there in the assets of the release. I don't know about TEDIT and UNICODE.

[MattHeffron]

Do you know about lispusers>fontsampler (which already had character sets as an option)?

I did not. I used FONTSAMPLE to check the BDF font conversion because you had recommended it to me back in December.
This can be closed without merging if that already has this functionality.

[masinter]

i'm confused about library/FONTSAMPLE vs. lispusers/FONTSAMPLER as they seem to be very different.
Also, neither one seems to print out an (enlarged) bitmap from the DISPLAYFONT -- I can't quite tell ...

Anyway, it would be good to at least have a description of what's going on.

[nbriggs]

Yes, they're different. I wrote FONTSAMPLER in 1985 because FONTSAMPLE didn't do what I wanted. Documentation on lispusers/FONTSAMPLER.TEDIT (FontSample (FONTCREATE 'CLASSIC 12) 239 NIL 'DISPLAY) will prompt you to sweep out a window in which it will display character set 239 of CLASSIC-12. You can give it a printer name and type if you want to create a Postscript or Interpress stream. FONTSAMPLE was expressly intended to create books of font samples on Interpress printers for the fonts known to Lisp and the printers. Documentation on library/FONTSAMPLE.TEDIT

[MattHeffron]

Yes, they're different. I wrote FONTSAMPLER in 1985 because FONTSAMPLE didn't do what I wanted.

Alas, it cannot create a PDF file showing the DISPLAYFONT glyphs (bitmaps), which would be really handy to check out my BDF to DISPLAYFONT conversion. So I added that ability: PR #2008 😉