4:37 p.m.
APPROVE 21.5 RECOMMEND 21.4
This patch optimises #'featurep for the cases...
(featurep 'xemacs) and (featurep 'sxemacs)
...making the former be always t and the latter be nil. It also works
for the advanced form of #'featurep...
(featurep '(and foo bar))
The advantage is that it will suppress byte-compiler warnings.
I recommend that this be added to 21.4 as well, although the regexp
needs to be converted from a raw string to a normal string (remove the
`#r' and double the number of '\').
Norbert, would you mind adding this to 21.5 for me. Thanks!
2007-09-02 Steve Youngs <steve(a)sxemacs.org>
* byte-optimize.el (byte-optimize-featurep): New. Optimise
feature 'sxemacs to nil and 'xemacs to t. This is along the same
lines as GNU/Emacs where they optimise both 'xemacs and 'sxemacs
to nil. It has the added advantage of suppressing byte-compiler
warnings in most cases.
--
|---<Steve Youngs>---------------<GnuPG KeyID: A94B3003>---|
| SXEmacs - The only _______ you'll ever need. |
| Fill in the blank, yes, it's THAT good! |
|------------------------------------<steve(a)sxemacs.org>---|
_______________________________________________
XEmacs-Patches mailing list
XEmacs-Patches(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-patches
10:51 a.m.
New subject: [Q] #'featurep byte-optimisation
Generally, the change seems fine. But do we really need to use regexps
for this kind of thing? It took me about 10 minutes to understand why
the regexp actually works. Can't we do something like this (untested
and probably at least a bit incorrect):
(defvar features-always-true '(xemacs sxemacs))
...
(defun byte-optimize-featurep (form)
(if (and (consp (cdr form))
(eq? 'quote (cadr form))
(consp (cddr form))
(memq (caddr form) features-always-true))
...))
--
Cheers =8-} Mike
Friede, Völkerverständigung und überhaupt blabla
_______________________________________________
XEmacs-Patches mailing list
XEmacs-Patches(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-patches
1:58 a.m.
New subject: [Q] #'featurep byte-optimisation
* Michael Sperber <sperber(a)informatik.uni-tuebingen.de> writes:
Generally, the change seems fine. But do we really need to use
regexps
for this kind of thing?
The regexp was the only way I could think of to have it work for both
the standard #'featurep use (featurep 'foo), and the
advanced... (featurep '(or foo bar))
(defvar features-always-true '(xemacs sxemacs))
...
(defun byte-optimize-featurep (form)
(if (and (consp (cdr form))
(eq? 'quote (cadr form))
(consp (cddr form))
(memq (caddr form) features-always-true))
...))
Wouldn't that return t even on XEmacs for (featurep 'sxemacs)?
--
|---<Steve Youngs>---------------<GnuPG KeyID: A94B3003>---|
| SXEmacs - The only _______ you'll ever need. |
| Fill in the blank, yes, it's THAT good! |
|------------------------------------<steve(a)sxemacs.org>---|
_______________________________________________
XEmacs-Patches mailing list
XEmacs-Patches(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-patches
8:49 a.m.
New subject: [Q] #'featurep byte-optimisation
Steve Youngs <steve(a)sxemacs.org> writes:
* Michael Sperber <sperber(a)informatik.uni-tuebingen.de>
writes:
> Generally, the change seems fine. But do we really need to use regexps
> for this kind of thing?
The regexp was the only way I could think of to have it work for both
the standard #'featurep use (featurep 'foo), and the
advanced... (featurep '(or foo bar))
Ah, I hadn't know about that. Does it do the right thing for, say,
(featurep '(and undefined-feature sxemacs))
? (My experiments suggest that it doesn't, but maybe I'm again not
understanding this right.)
> (defvar features-always-true '(xemacs sxemacs))
> ...
> (defun byte-optimize-featurep (form)
> (if (and (consp (cdr form))
> (eq? 'quote (cadr form))
> (consp (cddr form))
> (memq (caddr form) features-always-true))
> ...))
Wouldn't that return t even on XEmacs for (featurep 'sxemacs)?
Told you :-) I would think a small interpreter for the featurep language
is needed. I'll be happy to write it if it's the right thing to do.
--
Cheers =8-} Mike
Friede, Völkerverständigung und überhaupt blabla
_______________________________________________
XEmacs-Patches mailing list
XEmacs-Patches(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-patches
3:27 p.m.
New subject: [Q] #'featurep byte-optimisation
* Michael Sperber <sperber(a)informatik.uni-tuebingen.de> writes:
Steve Youngs <steve(a)sxemacs.org> writes:
> * Michael Sperber <sperber(a)informatik.uni-tuebingen.de> writes:
>
> > Generally, the change seems fine. But do we really need to use regexps
> > for this kind of thing?
Unfortunately, as Aidan pointed out, the patch was wrong anyway. I've
asked to have it reverted.
I would think a small interpreter for the featurep language is
needed. I'll be happy to write it if it's the right thing to do.
I'm not sure. I'm thinking now that the right thing to do is to leave
featurep as it was prior to me messing with it.
--
|---<Steve Youngs>---------------<GnuPG KeyID: A94B3003>---|
| SXEmacs - The only _______ you'll ever need. |
| Fill in the blank, yes, it's THAT good! |
|------------------------------------<steve(a)sxemacs.org>---|
_______________________________________________
XEmacs-Patches mailing list
XEmacs-Patches(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-patches
9:41 p.m.
New subject: [Q] #'featurep byte-optimisation
Ar an cúigiú lá de mí Méan Fómhair, scríobh Steve Youngs:
> I would think a small interpreter for the featurep language
is
> needed. I'll be happy to write it if it's the right thing to do.
I'm not sure. I'm thinking now that the right thing to do is to leave
featurep as it was prior to me messing with it.
Well, literally all of the #'featurep checks for XEmacs in the packages are
of the form:
(featurep 'xemacs)
because people, reasonably enough, want the expression to return nil instead
of provoking a wrong-type-argument on GNU. We don’t need anything complex to
optimise away *that* case. For the sake of differentiating our code from
GNU’s, though, until we get sucked into GPLV3 anyway, this is an
alternative:
(defvar byte-optimize-ever-present-features
;; Conservative list.
;; cl-19, cl-extras, backquote are also present from early in the dump
;; process, but they're not checked for anywhere in packages, nor in
;; Google code search.
'(xemacs cl))
(put 'featurep 'byte-optimizer 'byte-optimize-featurep)
(defun byte-optimize-featurep (form)
(if (memq (car-safe
(cdr-safe
(car-safe
(cdr-safe
form))))
byte-optimize-ever-present-features)
t
form))
--
On the quay of the little Black Sea port, where the rescued pair came once
more into contact with civilization, Dobrinton was bitten by a dog which was
assumed to be mad, though it may only have been indiscriminating. (Saki)
_______________________________________________
XEmacs-Patches mailing list
XEmacs-Patches(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-patches
1:57 p.m.
Ar an dara lá de mí Méan Fómhair, scríobh Steve Youngs:
APPROVE 21.5 RECOMMEND 21.4
This patch optimises #'featurep for the cases...
(featurep 'xemacs) and (featurep 'sxemacs)
...making the former be always t and the latter be nil.
The latter behaviour is inappropriate, isn’t it? Unless SXEmacs has moved to
only using packages that were compiled *by* SXEmacs; my understanding was
that it’s routine for files to be byte-compiled in XEmacs and run in
SXEmacs.
--
On the quay of the little Black Sea port, where the rescued pair came once
more into contact with civilization, Dobrinton was bitten by a dog which was
assumed to be mad, though it may only have been indiscriminating. (Saki)
_______________________________________________
XEmacs-Patches mailing list
XEmacs-Patches(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-patches
2 a.m.
* Aidan Kehoe <kehoea(a)parhasard.net> writes:
Ar an dara lá de mí Méan Fómhair, scríobh Steve Youngs:
> APPROVE 21.5 RECOMMEND 21.4
>
> This patch optimises #'featurep for the cases...
>
> (featurep 'xemacs) and (featurep 'sxemacs)
>
> ...making the former be always t and the latter be nil.
The latter behaviour is inappropriate, isn’t it?
Not at all. You don't want (featurep 'sxemacs) returning t on XEmacs do
you?
This same code in SXEmacs give t for both. The intention is so that you
_can_ write SXEmacs only code if you want to and not have problems with
it on XEmacs.
--
|---<Steve Youngs>---------------<GnuPG KeyID: A94B3003>---|
| SXEmacs - The only _______ you'll ever need. |
| Fill in the blank, yes, it's THAT good! |
|------------------------------------<steve(a)sxemacs.org>---|
_______________________________________________
XEmacs-Patches mailing list
XEmacs-Patches(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-patches
9:19 a.m.
Ar an ceathrú lá de mí Méan Fómhair, scríobh Steve Youngs:
* Aidan Kehoe <kehoea(a)parhasard.net> writes:
> Ar an dara lá de mí Méan Fómhair, scríobh Steve Youngs:
>> APPROVE 21.5 RECOMMEND 21.4
>>
>> This patch optimises #'featurep for the cases...
>>
>> (featurep 'xemacs) and (featurep 'sxemacs)
>>
>> ...making the former be always t and the latter be nil.
> The latter behaviour is inappropriate, isn’t it?
Not at all. You don't want (featurep 'sxemacs) returning t on XEmacs do
you?
Nope. But it shouldn’t be optimised away to nil if the code is to run on
SXEmacs, though! Test case, with your patch applied to XEmacs:
$ echo "(if (featurep 'sxemacs) (message \"in SXEmacs\") (message
\"not\"))" > /tmp/whatever.el
$ ~/xemacs-21.5/src/xemacs -batch -f batch-byte-compile /tmp/whatever.el
Compiling /tmp/whatever.el...
Wrote /tmp/whatever.elc
Done
$ ~/sxemacs-22.1.6/src/sxemacs -batch -l /tmp/whatever.elc
not
$
This same code in SXEmacs give t for both. The intention is so that
you
_can_ write SXEmacs only code if you want to and not have problems with
it on XEmacs.
--
On the quay of the little Black Sea port, where the rescued pair came once
more into contact with civilization, Dobrinton was bitten by a dog which was
assumed to be mad, though it may only have been indiscriminating. (Saki)
_______________________________________________
XEmacs-Patches mailing list
XEmacs-Patches(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-patches
1:57 p.m.
* Aidan Kehoe <kehoea(a)parhasard.net> writes:
Ar an ceathrú lá de mí Méan Fómhair, scríobh Steve Youngs:
> * Aidan Kehoe <kehoea(a)parhasard.net> writes:
>
> > Ar an dara lá de mí Méan Fómhair, scríobh Steve Youngs:
> >> APPROVE 21.5 RECOMMEND 21.4
> >>
> >> This patch optimises #'featurep for the cases...
> >>
> >> (featurep 'xemacs) and (featurep 'sxemacs)
> >>
> >> ...making the former be always t and the latter be nil.
>
> > The latter behaviour is inappropriate, isn’t it?
>
> Not at all. You don't want (featurep 'sxemacs) returning t on XEmacs do
> you?
Nope. But it shouldn’t be optimised away to nil if the code is to run
on
SXEmacs, though! Test case, with your patch applied to XEmacs:
$ echo "(if (featurep 'sxemacs) (message \"in
SXEmacs\") (message \"not\"))" > /tmp/whatever.el
$ ~/xemacs-21.5/src/xemacs -batch -f batch-byte-compile /tmp/whatever.el
Compiling /tmp/whatever.el...
Wrote /tmp/whatever.elc
Done
$ ~/sxemacs-22.1.6/src/sxemacs -batch -l /tmp/whatever.elc
not
$
Oh hell! You are absolutely right. And it would be just as bad going
the other way... compile with SXE and run with XE.
Here's the patch to revert it, could somebody please apply it for me
(Norbert?)
2007-09-04 Steve Youngs <steve(a)sxemacs.org>
* byte-optimize.el (byte-optimize-featurep): Remove. It's wrong.
--
|---<Steve Youngs>---------------<GnuPG KeyID: A94B3003>---|
| SXEmacs - The only _______ you'll ever need. |
| Fill in the blank, yes, it's THAT good! |
|------------------------------------<steve(a)sxemacs.org>---|
_______________________________________________
XEmacs-Patches mailing list
XEmacs-Patches(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-patches
6318
days inactive
6321
days old
9 comments
3 participants
participants (3)
-
Aidan Kehoe -
Michael Sperber -
Steve Youngs