bug#11594: message-send-and-exit now fails

bug#11594: message-send-and-exit now fails

[Jim Meyering]

I built emacs from the latest on the trunk and found
that I can no longer send mail in the usual manner (using gnus).

It worked fine with a snapshot from 2012-05-24.09h10 UTC.
Here's how it fails, now:

Debugger entered--Lisp error: (wrong-type-argument number-or-marker-p nil)
  auth-source-search(:host "localhost" :port "smtp" :user nil :max 1 :require nil :create nil)
  smtpmail-try-auth-methods(#<process smtpmail> (dsn 8bitmime enhancedstatuscodes etrn size) "localhost" "smtp" nil)
  byte-code("...elided..." [host process-buffer buffer-undo-list coding-system-for-write coding-system-for-read port get-buffer-create format "*trace of SMTP session to %s*" t erase-buffer binary open-network-stream "smtpmail" :type :return-list :capability-command "EHLO %s.\n" smtpmail-fqdn :end-of-command "^[0-9]+ .*.\n" :success "^2.*\n" :always-query-capabilities :starttls-function #[(capabilities) "\301\302\b\"\205\b.\303\207" [capabilities string-match "[ -]STARTTLS" "STARTTLS.\n"] 3] :client-certificate :use-starttls-if-possible throw done plist-get :error "Unable to contact server" set-process-filter smtpmail-process-filter :greeting smtpmail-response-code "No greeting: %s" 400 "Connection not allowed: %s" set-buffer-process-coding-system raw-text-unix make-local-variable smtpmail-re
 ad-point :capabilities smtpmail-command-or-throw "HELO %s" delete "" split-string ...] 24)
  smtpmail-via-smtp(("11557@debbugs.gnu.org" "monnier@iro.umontreal.ca" "rgm@gnu.org") #<buffer  smtpmail temp>)
  smtpmail-send-it()
  message-smtpmail-send-it()
  #[nil "\b \207" [message-send-mail-function] 1]()
  gnus-agent-send-mail()
  message-send-mail(nil)
  message-send-via-mail(nil)
  message-send(nil)
  message-send-and-exit(nil)
  call-interactively(message-send-and-exit nil nil)

bug#11594: message-send-and-exit now fails

[Achim Gratz]

Jim Meyering writes:
> I built emacs from the latest on the trunk and found
> that I can no longer send mail in the usual manner (using gnus).

Actually, Gnus is completely hosed and doesn't even start up for me and
it appears that the auth code is involved (as I could not find any
suspect changes in Gnus between the last working version and yesterdays
pull).


Regards,
Achim.
-- 
+<[Q+ Matrix-12 WAVE#46+305 Neuron microQkb Andromeda XTk Blofeld]>+

Factory and User Sound Singles for Waldorf Blofeld:
http://Synth.Stromeko.net/Downloads.html#WaldorfSounds

bug#11594: message-send-and-exit now fails

[Johan Bockgård]

Achim Gratz <Stromeko@nexgo.de> writes:

> Jim Meyering writes:
>> I built emacs from the latest on the trunk and found
>> that I can no longer send mail in the usual manner (using gnus).
>
> Actually, Gnus is completely hosed and doesn't even start up for me and
> it appears that the auth code is involved (as I could not find any
> suspect changes in Gnus between the last working version and yesterdays
> pull).

The byte compiler seems broken:

    emacs -Q

    (setq m (byte-compile '(lambda () (defmacro m ()) (m))))

    #[nil "\300\301\302\303B\"\210\301 \207" [defalias m macro #[nil "\300\207" [nil] 1]] 4]

    (funcall m)

       => error: (invalid-function m)


In emacs-24 branch:

    #[nil "\300\301\302\"\210\303\207" [defalias m (macro . #[nil "\300\207" [nil] 1]) nil] 3]

bug#11594: message-send-and-exit now fails

[Andreas Schwab]

Johan Bockgård <bojohan@gnu.org> writes:

> The byte compiler seems broken:
>
>     emacs -Q
>
>     (setq m (byte-compile '(lambda () (defmacro m ()) (m))))
>
>     #[nil "\300\301\302\303B\"\210\301 \207" [defalias m macro #[nil "\300\207" [nil] 1]] 4]
>
>     (funcall m)
>
>        => error: (invalid-function m)

byte code:
  args: nil
0	constant  defalias
1	constant  m
2	constant  macro
3	constant  <compiled-function>
      args: nil
    0	    constant  nil
    1	    return    

4	cons	  
5	call	  2
6	discard	  
7	constant  m
8	call	  0
9	return	  

> In emacs-24 branch:
>
>     #[nil "\300\301\302\"\210\303\207" [defalias m (macro . #[nil "\300\207" [nil] 1]) nil] 3]

byte code:
  args: nil
0	constant  defalias
1	constant  m
2	constant  <compiled macro>
      args: nil
    0	    constant  nil
    1	    return    

3	call	  2
4	discard	  
5	constant  nil
6	return	  

There are two differences:

- (cons 'macro #[nil ...]) vs '(macro . #[nil ...]) (a performance
  regression)
- (m) is not macro-expanded

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#11594: message-send-and-exit now fails

[Johan Bockgård]

Andreas Schwab <schwab@linux-m68k.org> writes:

> There are two differences:
>
> - (cons 'macro #[nil ...]) vs '(macro . #[nil ...]) (a performance
>   regression)
> - (m) is not macro-expanded

Actually, this doesn't seem to be responsible.


I think the bug is in this part:

2012-05-30  Stefan Monnier  <monnier@iro.umontreal.ca>

...
        (byte-compile-output-file-form): Move print-vars binding.
        (byte-compile-output-docform): Simplify accordingly.

bug#11594: message-send-and-exit now fails

[Johan Bockgård]

Johan Bockgård <bojohan@gnu.org> writes:

> I think the bug is in this part:
>
> 2012-05-30  Stefan Monnier  <monnier@iro.umontreal.ca>
>
> ...
>         (byte-compile-output-file-form): Move print-vars binding.
>         (byte-compile-output-docform): Simplify accordingly.

... since the #N= stuff seems to be mostly missing in the elc files now.

bug#11594: message-send-and-exit now fails

[Stefan Monnier]

> There are two differences:

> - (cons 'macro #[nil ...]) vs '(macro . #[nil ...]) (a performance
>   regression)
> - (m) is not macro-expanded

Actually, I think his tests are buggy.  When I try it on Emacs-24 and
Emacs-23, the macro is not expanded either.  As for the performance,
the difference only applies to non-toplevel defmacros.


        Stefan

bug#11594: message-send-and-exit now fails

[Johan Bockgård]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> There are two differences:
>
>> - (cons 'macro #[nil ...]) vs '(macro . #[nil ...]) (a performance
>>   regression)
>> - (m) is not macro-expanded
>
> Actually, I think his tests are buggy.

Yes, I realized this. What causes the problems is that print-circle is
not bound correctly. See my other posts in this thread.

bug#11594: message-send-and-exit now fails

[Stefan Monnier]

> Yes, I realized this.  What causes the problems is that print-circle is
> not bound correctly.  See my other posts in this thread.

Indeed, eliminating print-circle from byte-compile-output-docform was
correct when I did it (at which point -defmumble did not call
byte-compile-output-docform any more), but later I re-added the call to
byte-compile-output-docform and forgot to re-add the corresponding
print-circle.  Thanks for helping me catch it, should be fixed now,


        Stefan