Subject: Re: bogus change(s) in cl-macs.el - msg#00024

List: emacs.w3m

Date: Prev Next Index Thread: Prev Next Index

> First of all, I protest strongly that such a big change has not
> been recorded on the ChangeLog file.

The ChangeLog is somewhere, but as usual I forgot to commit it at the time
and now I'm working on another computer. But trust me, it's not lost.

> *** cl-macs.el~ Tue Nov 16 21:54:14 2004
> --- cl-macs.el Fri Nov 19 01:30:39 2004
> ***************
> *** 1314,1320 ****
> \(fn ((FUNC ARGLIST BODY...) ...) FORM...)"
> (let ((vars nil) (sets nil) (cl-macro-environment cl-macro-environment))
> (while bindings
> ! (let ((var (make-symbol "--cl-var--")))
> (push var vars)
> (push (list 'function* (cons 'lambda (cdar bindings))) sets)
> (push var sets)
> --- 1314,1320 ----
> \(fn ((FUNC ARGLIST BODY...) ...) FORM...)"
> (let ((vars nil) (sets nil) (cl-macro-environment cl-macro-environment))
> (while bindings
> ! (let ((var (gensym)))
> (push var vars)
> (push (list 'function* (cons 'lambda (cdar bindings))) sets)
> (push var sets)

It's odd that such a change would fix/break some code.
After all `gensym' is more or less doing (make-symbol (concat "G"
(random))), so other than the symbol's name, nothing is changed.
And since the symbol is uninterned, its name is mostly irrelevant, unless
some code later on uses `symbol-name' or does a print+read without using
print-gensym.

Maybe we'll indeed have to revert to gensym here, but I'd like to first
better understand the problem.

Could you give us a recipe to reproduce the problem, and a backtrace
if applicable. Of course, ideally the recipe should not use emacs-w3m,
but as a first step it would still help to see a recipe and a description of
the resulting problem.

Stefan

Was this page helpful?

Thread at a glance:

Previous Message by Date: click to view message preview

Re: bogus change(s) in cl-macs.el

> I myself suggested the following alternate type solution, which I > believe would not have broken anything: > > ! (let ((var (with-no-warnings (gensym)))) As well as being unclean, that doesn't solve the problem of gensym being undefined at runtime. `gensym' _only_ needs to be defined at runtime when gensym is going to be _called_ at runtime. I said that I still had to double check my alternative patch should it be needed. I did check in the `winner' case that `gensym' was _not_ called at runtime. I would strongly guess that this is the case with _all_ these calls to `gensym'. Clearly, I would have to _check_ that this is really the case. All of this only matters if there are real problems with Stefan's patch. Sincerely, Luc.

Next Message by Date: click to view message preview

[emacs-w3m:07185] Re: bogus change(s) in cl-macs.el

>>>>>> In [emacs-w3m : No.07180] Miles Bader wrote: >> Please describe this "breakage" in greater detail. >>>>> In [emacs-w3m : No.07182] Katsumi Yamaoka wrote: > From yesterday I'm looking for a simpler way rather than to use > emacs-w3m to reproduce the bug. However, it has not been > achieved so far. I'll post it when I found it. I found the simplest way to explain the bug: (macroexpand '(labels ((FOO nil FOO-BODY) (BAR nil BAR-BODY)) (FOO) (BAR))) (let ((--cl---cl-var---- nil) (--cl---cl-var---- nil)) (progn (progn (set '--cl---cl-var---- #'(lambda nil BAR-BODY)) (set '--cl---cl-var---- #'(lambda nil FOO-BODY))) (funcall (symbol-value '--cl---cl-var----)) (funcall (symbol-value '--cl---cl-var----)))) One of two functions FOO and BAR is disregarded as you see. When the patch I posted is applied, it will be corrected as follows: (let ((--cl-G79813-- nil) (--cl-G79812-- nil)) (progn (progn (set '--cl-G79813-- #'(lambda nil BAR-BODY)) (set '--cl-G79812-- #'(lambda nil FOO-BODY))) (funcall (symbol-value '--cl-G79812--)) (funcall (symbol-value '--cl-G79813--))))

Previous Message by Thread: click to view message preview

Re: bogus change(s) in cl-macs.el

> Well, actually we can't see it here because you haven't used print-gensym > to distinguish symbols with the same name. But indeed: > (labels ((foo nil 1) (bar nil 2)) (cons (foo) (bar))) > returns (1 . 1). This is because cl's macro environment uses symbol names > rather than symbols for symbol-macros, so if two symbols have the same name > (i.e. `eq', not just `equal') and they both are symbol-macros, then > CL screws up. I'm working on a patch for that. OK, I've committed basically your fix (except it also needed to be applied at one other place, and I added some comment explaining why we use `gensym' rather than `make-symbol' there). Stefan

Next Message by Thread: click to view message preview

[emacs-w3m:07195] antenna supports RSS

RSS で更新が告知されているサイトを，antenna で補足できるようにしてみました．・w3m-rss.el と sb-rss.el でかなりコードが重複してしまっているので整理が必要．・最初のチャンネルの更新日時だけを使っているので，チャンネルの更新日時が指定されていない場合はうまく動かない．などの問題があります．とりあえずの叩き台ですので，適当に改良して頂けるとありがたいです． -- 土屋雅稔 ( TSUCHIYA Masatoshi )

Loading Comments...