mailcap - caching "nil" failed test

mailcap - caching "nil" failed test

[Kevin Ryde]

[-- Attachment #1: Type: text/plain, Size: 810 bytes --]

If there's a mailcap-mime-data entry with no 'test clause, then I
think mailcap-viewer-test-cache is getting an entry "(nil t)" which
means that "nil" for failure comes back as "t" for success.

I struck this with a silly mozilla entry in the debian /etc/mailcap,
"text/html; false; x-mozilla-flags=internal" when viewing some html.
It then makes failed tests like my pdf on a dumb tty example end up
getting run.

2006-01-28  Kevin Ryde  <user42@zip.com.au>

	* mailcap.el (mailcap-viewer-passes-test): Don't put "(nil t)" into
	mailcap-viewer-test-cache when there's no 'test clause, since that
	will invert the meaning of a "nil" test previously determined by
	mailcap-mailcap-entry-passes-test.

In the diff below I didn't reindent the second half of the func, so as
to best show the actual bits changed.


[-- Attachment #2: mailcap.el.no-cache.diff --]
[-- Type: text/plain, Size: 981 bytes --]

--- mailcap.el.~7.8.~	2006-01-26 09:44:03.000000000 +1100
+++ mailcap.el	2006-01-28 09:01:39.000000000 +1100
@@ -644,12 +644,13 @@
 	 (viewer (cdr (assoc 'viewer viewer-info)))
 	 (default-directory (expand-file-name "~/"))
 	 status parsed-test cache result)
-    (if (setq cache (assoc test mailcap-viewer-test-cache))
-	(cadr cache)
+    (cond ((setq cache (assoc test mailcap-viewer-test-cache))
+	   (cadr cache))
+          ((not test-info) t)		; No test clause
+          (t
       (setq
        result
        (cond
-	((not test-info) t)		; No test clause
 	((not test) nil)		; Already failed test
 	((eq test t) t)			; Already passed test
 	((functionp test)		; Lisp function as test
@@ -667,7 +668,7 @@
 	       status (apply 'call-process test))
 	 (eq 0 status))))
       (push (list otest result) mailcap-viewer-test-cache)
-      result)))
+      result))))
 
 (defun mailcap-add-mailcap-entry (major minor info)
   (let ((old-major (assoc major mailcap-mime-data)))

Re: mailcap - caching "nil" failed test

[Katsumi Yamaoka]

>>>>> In <87mzhhjp1e.fsf@zip.com.au> Kevin Ryde wrote:

> If there's a mailcap-mime-data entry with no 'test clause, then I
> think mailcap-viewer-test-cache is getting an entry "(nil t)" which
> means that "nil" for failure comes back as "t" for success.

> I struck this with a silly mozilla entry in the debian /etc/mailcap,
> "text/html; false; x-mozilla-flags=internal" when viewing some html.
> It then makes failed tests like my pdf on a dumb tty example end up
> getting run.

> 2006-01-28  Kevin Ryde  <user42@zip.com.au>

> 	* mailcap.el (mailcap-viewer-passes-test): Don't put "(nil t)" into
> 	mailcap-viewer-test-cache when there's no 'test clause, since that
> 	will invert the meaning of a "nil" test previously determined by
> 	mailcap-mailcap-entry-passes-test.

> In the diff below I didn't reindent the second half of the func, so as
> to best show the actual bits changed.

I've applied your patch to the trunk and the v5-10 branch.
Thank you.

Re: mailcap - caching "nil" failed test

[Zlatko Calusic]

Katsumi Yamaoka <yamaoka@jpl.org> writes:

>>>>>> In <87mzhhjp1e.fsf@zip.com.au> Kevin Ryde wrote:
>
>> If there's a mailcap-mime-data entry with no 'test clause, then I
>> think mailcap-viewer-test-cache is getting an entry "(nil t)" which
>> means that "nil" for failure comes back as "t" for success.
>
>> I struck this with a silly mozilla entry in the debian /etc/mailcap,
>> "text/html; false; x-mozilla-flags=internal" when viewing some html.
>> It then makes failed tests like my pdf on a dumb tty example end up
>> getting run.
>
>> 2006-01-28  Kevin Ryde  <user42@zip.com.au>
>
>> 	* mailcap.el (mailcap-viewer-passes-test): Don't put "(nil t)" into
>> 	mailcap-viewer-test-cache when there's no 'test clause, since that
>> 	will invert the meaning of a "nil" test previously determined by
>> 	mailcap-mailcap-entry-passes-test.
>
>> In the diff below I didn't reindent the second half of the func, so as
>> to best show the actual bits changed.
>
> I've applied your patch to the trunk and the v5-10 branch.
> Thank you.
>

Hm, I don't know if its connected to this (and the second one) change
to the mailcap.el, but recently I have started having problems when
clicking on attachments with spaces in the name. It was working nice
for the last year or so, and now it once again stopped working. The
system is Debian Linux unstable with gnus 5.10.6-1.NO.20051213-1
(watch out, version number is misleading!).

Is this a recession in the trunk or should I report this as a debian
bug?
-- 
Zlatko

Re: mailcap - caching "nil" failed test

[Kevin Ryde]

Zlatko Calusic <zlatko.calusic@iskon.hr> writes:
>
> Hm, I don't know if its connected to this (and the second one) change
> to the mailcap.el, but recently I have started having problems when
> clicking on attachments with spaces in the name. It was working nice
> for the last year or so, and now it once again stopped working. The
> system is Debian Linux unstable with gnus 5.10.6-1.NO.20051213-1
> (watch out, version number is misleading!).

The only thing my change should have done (hopefully) is fix which
entries get used.  Unless you've got some dodgy entries now being
picked up then it oughtn't have hurt anything.