From: "Dr. Arne Babenhauserheide" <arne_bab@web.de>
To: tomas@tuxteam.de
Cc: 62053@debbugs.gnu.org
Subject: bug#62053: Guile compiler is unable to compile tiny program
Date: Wed, 08 Mar 2023 18:38:10 +0100 [thread overview]
Message-ID: <87bkl36suw.fsf@web.de> (raw)
In-Reply-To: <ZAi6r0pSjEbrMJiD@tuxteam.de>
[-- Attachment #1: Type: text/plain, Size: 2437 bytes --]
<tomas@tuxteam.de> writes:
>> (define (fn l)
>> (define (ok l)
>> (or (and (pair? l) l)
>> (error "list is empty")))
>> (cons (reverse (car l)) (ok l)))
> (so in my case it's 25 instead of 22). Interestingly, my version also
> produces warnings about possibly unbound variables). If I enter the
> code line by line into a repl, I also get a "not found 22".
>
> Interestingly, if I add a call to fn to the end of that snippet
> (e.g. (fn '(a b))), the "not found" disappears. The unbound var
> warnings stay.
Also interesting: I can primitive-eval the code:
(define code '(define (fn l)
(define (ok l)
(or (and (pair? l) l)
(error "list is empty")))
(cons (reverse (car l)) (ok l))))
(primitive-eval code)
(fn '((1 2))) ;; => ((2 1) (1 2))
What might be happening here is control flow analysis discarding
impossible branches:
(car l) will throw an error when l is not a pair. Since arguments are
evaluated in order, (ok l) will always succeed. This means (ok l) can be
replaced by l.
To look into details:
(compile code) ;; => throws an error: not found 22
;; use ,bt to get:
,bt
In system/base/compile.scm:
352:28 17 (compile _ #:from _ #:to _ #:env _ #:optimization-level _ #:warning-level _ #:opts _)
265:44 16 (_ _ _)
265:44 15 (_ _ _)
261:33 14 (_ #<intmap 0-83> #<module (#{ g836}#) 7fcbf85dfbe0>)
In language/cps/optimize.scm:
130:12 13 (_ _ _)
102:3 12 (optimize-higher-order-cps _ _)
In language/cps/type-fold.scm:
775:16 11 (type-fold _)
In language/cps/renumber.scm:
170:36 10 (renumber #<intmap 0-4,6-21,52-53> _)
166:30 9 (compute-renaming _ _)
144:7 8 (visit-fun _ _ _)
160:19 7 (visit-fun 12 #<transient-intmap 0-4,6-11> #<transient-intmap 0-3,5-6>)
In language/cps/utils.scm:
245:39 6 (compute-predecessors #<intmap 0-4,6-21,52-53> _ #:labels _)
107:3 5 (compute-function-body _ _)
129:30 4 (visit-cont _ _)
129:30 3 (visit-cont _ _)
112:9 2 (visit-cont 22 _)
In language/cps/intmap.scm:
397:51 1 (_ _)
In ice-9/boot-9.scm:
1685:16 0 (raise-exception _ #:continuable? _)
I’ll leave further steps to folks who are more experienced with
optimize-higher-order-cps (or want to become more experienced with it).
Best wishes,
Arne
--
Unpolitisch sein
heißt politisch sein,
ohne es zu merken.
draketo.de
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 1125 bytes --]
prev parent reply other threads:[~2023-03-08 17:38 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-03-08 11:07 bug#62053: Guile compiler is unable to compile tiny program var-vniiaes--- via Bug reports for GUILE, GNU's Ubiquitous Extension Language
2023-03-08 16:41 ` tomas
2023-03-08 17:38 ` Dr. Arne Babenhauserheide [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
List information: https://www.gnu.org/software/guile/
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=87bkl36suw.fsf@web.de \
--to=arne_bab@web.de \
--cc=62053@debbugs.gnu.org \
--cc=tomas@tuxteam.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).