"Unrecognised file info key" %:PACKAGE

[RyanBurnside]

I've Lisped myself into a pickle it seems.
I did reinstall Medley and run the suggested scripts in our last meeting. (seems good now)

I'm currently trying to save updated function definitions in some files and I get the following message.

I've got my image saved so I can reopen it but I'd really like to get these files fixed.
Not sure if loading LOOPS for my project causes this or not.
I also may have tried to CL:COMPILE-FILE at some point (to see if there was a speed increase) BUT that was a few sessions ago.

Any help appreciated. My Boids demo is mostly done now with a few more little functions here and there.

[nbriggs]

Given that it's failing in PRETTYDEF, I'd try (MAKEFILE 'BOIDS '(NEW FAST)) and see if it manages to do it without going through PRETTYDEF. NEW means don't try and save printing by copying definitions from the last version of the file, just use what's in memory, and FAST means don't bother pretty-printing on the disk file.

[RyanBurnside]

I gave it a shot ...

Oddly enough, PRETTYDEF appears in the new backtrace.

[nbriggs]

Yeah, OK -- it's coming from \DO-DEFINE-FILE-INFO (on sources/BOOTSTRAP). @rmkaplan was last one in there.

[nbriggs]

Try (in an interlisp exec) (TRACE \DO-DEFINE-FILE-INFO) and then try the MAKEFILE again -- you're looking for something odd, such as the list of define-file-info parameter keys having something in a peculiar package, so note the value of *PACKAGE* in the break window.

[RyanBurnside]

From what I see the ARGS don't match between first and 3rd lines.
The keywords get prefixed by %.

[nbriggs]

Yes, but the code in \DO-DEFINE-FILE-INFO reads

(SELECTQ (CAR TAIL)
       ((:PACKAGE %:PACKAGE)
        (SETQ PACKAGE (if (LISTP VALUE)
                          then
                          (SETQ PACKAGEFORM VALUE)
                          (EVAL VALUE)
                          ELSE VALUE))
        (IF (TYPEP PACKAGE 'PACKAGE)
            ELSEIF
            (SETQ PACKAGE (CL:FIND-PACKAGE PACKAGE))
            ELSE
            
            (* ;; "Better message than just \DTEST")

            (ERROR "Can't find package for DEFINE-FILE-INFO reader environment" VALUE)))
       ((:READTABLE %:READTABLE)
        (SETQ READTABLE (if (LISTP VALUE)
                            then
                            (SETQ READTABLEFORM VALUE)
                            (EVAL VALUE)
                            ELSE VALUE))
        (IF (TYPEP READTABLE 'READTABLEP)
            ELSEIF
            (SETQ READTABLE (FIND-READTABLE READTABLE))
            ELSE
            
            (* ;; "Better message than just \DTEST")

            (ERROR "Can't find read table for DEFINE-FILE-INFO reader environment" VALUE)))
       ((:BASE %:BASE)
                                                             (* ; 
                                                          "RMK:  An EVAL form here makes no sense.  ")
        (SETQ BASE (OR (\CHECKRADIX (if (LISTP VALUE)
                                        then
                                        (EVAL VALUE)
                                        else VALUE))
                       (ERROR "Bad read base for DEFINE-FILE-INFO reader environment" VALUE))))
       ((:FORMAT FORMAT %:FORMAT)
        (SETQ FORMAT (FETCH (EXTERNALFORMAT NAME)
                            OF
                            (FIND-FORMAT VALUE))))
       (ERROR "Unrecognized file info key" (CAR TAIL)))

So it's supposed to catch the incoming package keyword when it's actually IL:\:PACKAGE (as it would appear if printed in an XCL exec that was in the XCL-USER package by default)

[nbriggs]

Could you make available (attach, send a link, whatever) the last good version of the BOIDS file? I'd like to see if there's something odd about the DEFINE-FILE-INFO at the beginning. Also, what is the IL:MAKESYSDATE in your running image, and what is the (Interlisp exec) (GETPROP 'BOOTSTRAP 'FILEDATES) so that I can identify what code it's running.

[RyanBurnside]

Sure.
BOIDS depends on RBMATH and also LOOPS.
I usually load these manually before calling (BOIDWORLD) from Boids.

I've included BOIDS and RBMATH but you'll probably need LOOPS if you try to run my demo.
I usually load RBMATH, LOOPS, then BOIDS.

Ryan Files.zip

[RyanBurnside]

Here is the vmem file. I dunno how portable it is...
lisp.virtualmem.zip

[nbriggs]

Portable enough - I renamed the lisp.virtualmem to boids.sysout and then started that running. Connected to {dsk}/tmp (because I have different directory structure) and did the (MAKEFILE 'BOIDS) - got the error, typed REVERT \DO-DEFINE-FILE-INFO in the break window, then (SEDIT::SEDIT ARGS) and edit the %:... to just :..., quit SEdit, and type OK to the break. Wrote the file out fine. What I can't tell you is why it failed. I would be suspicious of a LOOPS interaction though. It looked as though you also had changes to RBMATH - do the same trick.

[nbriggs]

Somehow... the MAKEFILE-ENVIRONMENT property of IL:BOIDS (the atom that's the base name of the file) has got the %:xxx etc. atoms instead of the :xxx etc. atoms. That's where PRETTYDEF is getting the file info it doesn't like. I don't see that on the last disk file you wrote - it looks fine there, so I can't explain what happened. You can fix it directly on the property on BOIDS, and do the MAKEFILE. I'd suggest reloading the file into a new sysout. @rmkaplan @masinter -- any ideas on what might have happened?

[pamoroso]

I don't know whether it's related but in LOOPS I've had issues with symbols containing the : character.

Some of the instance variables of my GravityLoops program have the property :initForm. I initially typed it literally in SEdit but got errors when saving. Escaping each : with % like %:initForm resolved the issue and I just asumed LOOPS has its own readtable.

[rmkaplan]

In PRETTYDEF, it looks a little odd that the code that encloses the first call to \DO-DEFINE-FILE-INFO is setting the DESTINATIONENV instead of the SOURCENV, unlike the other branch. That may or may not have anything to do with this. (if (SETQ DESTINATIONENV (GET ROOTNAME 'MAKEFILE-ENVIRONMENT)) then (SETQ DESTINATIONENV (\DO-DEFINE-FILE-INFO NIL DESTINATIONENV)) else (* ; "see if we already know the environment of the source") (CL:MULTIPLE-VALUE-SETQ (SOURCEFILENV OLDFILEMAP SOURCEFC) (LOOKUP-ENVIRONMENT-AND-FILEMAP (OR SOURCEFILE ROOTNAME) (OR (NULL SOURCEFILE) (EQ SOURCEFILE ROOTNAME)))))

…

On Jun 15, 2026, at 7:11 AM, Nick Briggs ***@***.***> wrote: Somehow... the MAKEFILE-ENVIRONMENT property of IL:BOIDS (the atom that's the base name of the file) has got the %:xxx etc. atoms instead of the :xxx etc. atoms. That's where PRETTYDEF is getting the file info it doesn't like. I don't see that on the last disk file you wrote - it looks fine there, so I can't explain what happened. You can fix it directly on the property on BOIDS, and do the MAKEFILE. I'd suggest reloading the file into a new sysout. @rmkaplan <https://github.com/rmkaplan> @masinter <https://github.com/masinter> -- any ideas on what might have happened? — Reply to this email directly, view it on GitHub <#2654?email_source=notifications&email_token=AQSTUJPLT7VCHEKBBCUVQFT4777YBA5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCNZTGA4TGMJTUZZGKYLTN5XKO3LFNZ2GS33OUVSXMZLOOSWGM33PORSXEX3DNRUWG2Y#discussioncomment-17309313>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AQSTUJMSMUFQ5KVC6XHUPRT4777YBAVCNFSNUABIKJSXA33TNF2G64TZHMZDQMJVGE4TSNRUHNCGS43DOVZXG2LPNY5TCMBSGU3TGOBWUF3AE>. You are receiving this because you were mentioned.

[nbriggs]

The problem seems to be that the BOOTSTRAP.LCOM does not reflect the source code in BOOTSTRAP for \DO-DEFINE-FILE-INFO -- it has two checks for :keyword while the source has :keyword and %:keyword. Looks like a recompile failure?

[nbriggs]

I've just verified that loading source for only \DO-DEFINE-FILE-INFO and recompiling it in Ryan's lisp.virtualmem solves the problem of being able to MAKEFILE his BOIDS file. I still have no idea how the MAKEFILE-ENVIRONMENT property on his file ended up with the %:xxx keys - but since there was an accommodation put in for that, it's not unexpected?

[RyanBurnside]

I tried loading the source in the editor and compiling it in SEDIT with ctrl c.
Looks to have given a "SIMPLE ERROR".

[nbriggs]

It requires exports.all be loaded to compile that code. BUT... we've fixed it in the repo in medley/sources/BOOTSTRAP.LCOM, if you check out the branch rmk198--Recompile-DO-DEFINE-FILE-INFO-in-BOOTSTRAP, then

179_ CONN {MEDLEY}<sources>
{MEDLEY}<sources>
180_ (LOADFNS '\DO-DEFINE-FILE-INFO "BOOTSTRAP.LCOM")

(\DO-DEFINE-FILE-INFO redefined)
(\DO-DEFINE-FILE-INFO)
181_ CONN
{DSK}<home>ryan>il>
182_ (MAKEFILE 'BOIDS)
{DSK}<home>ryan>il>BOIDS..... (not sure what version it'll be)
183_

[RyanBurnside]

I'll give it a shot, I appreciate the time spent discussing this.
EDIT, took some tinkering to get compiled and loaded from the branch but it worked!

Thanks!