Emacs-24.5-1-universal.dmg for macosx dumps on first keystroke

Emacs-24.5-1-universal.dmg for macosx dumps on first keystroke

[Randy Bush]

13" mid 2014 macbook pro
10.11.6 (15G31)
XQuartz 2.7.9 (xorg-server 1.17.4)
Emacs-24.5-1-universal.dmg

started either by
  /Applications/Emacs.app/Contents/MacOS/Emacs at bash cli within an   xterm
  or by double clicking the icon

with or without ~/.emacs

immediately after start, hit C-h to try to find the value of
window-system.  crash.

what macos gives me for the crash is at
https://archive.psg.com/emacs24.trace

any clues on how to further debug much appreciated.

fwiw, this email written in wanderlust under 
   GNU Emacs 22.3.1 (i386-apple-darwin9.8.0, Carbon Version 1.6.0) of 2010-01-10 on gs674-seijiz.local
which i have been running happily for years

randy

Emacs-24.5-1-universal.dmg for macosx dumps on first keystroke

[Randy Bush]

13" mid 2014 macbook pro
10.11.6 (15G31)
XQuartz 2.7.9 (xorg-server 1.17.4)
Emacs-24.5-1-universal.dmg

started either by
  /Applications/Emacs.app/Contents/MacOS/Emacs at bash cli within an   xterm
  or by double clicking the icon

with or without ~/.emacs

immediately after start, hit C-h to try to find the value of
window-system.  crash.

what macos gives me for the crash is at
https://archive.psg.com/emacs24.trace

any clues on how to further debug much appreciated.

fwiw, this email written in wanderlust under 
   GNU Emacs 22.3.1 (i386-apple-darwin9.8.0, Carbon Version 1.6.0) of 2010-01-10 on gs674-seijiz.local
which i have been running happily for years

randy

Re: Emacs-24.5-1-universal.dmg for macosx dumps on first keystroke

[Alan Third]

Hi,

I'd recommend you try the 25.1-rc2 release as there have been a lot of
crashes fixed.



On 25 August 2016 at 10:04, Randy Bush <randy@psg.com> wrote:
> 13" mid 2014 macbook pro
> 10.11.6 (15G31)
> XQuartz 2.7.9 (xorg-server 1.17.4)
> Emacs-24.5-1-universal.dmg
>
> started either by
>   /Applications/Emacs.app/Contents/MacOS/Emacs at bash cli within an   xterm
>   or by double clicking the icon
>
> with or without ~/.emacs
>
> immediately after start, hit C-h to try to find the value of
> window-system.  crash.
>
> what macos gives me for the crash is at
> https://archive.psg.com/emacs24.trace
>
> any clues on how to further debug much appreciated.
>
> fwiw, this email written in wanderlust under
>    GNU Emacs 22.3.1 (i386-apple-darwin9.8.0, Carbon Version 1.6.0) of 2010-01-10 on gs674-seijiz.local
> which i have been running happily for years
>
> randy
>



-- 
Alan Third

Re: Emacs-24.5-1-universal.dmg for macosx dumps on first keystroke

[Randy Bush]

> I'd recommend you try the 25.1-rc2 release as there have been a lot of
> crashes fixed.

the crash dump from Emacs-pretest-25.1-rc2-universal.dmg is at
https://archive.psg.com/emacs25.1-rc2.trace

the crash dump from
Emacs-2016-08-25_01-41-34-70cfe9df957bd2f0699f62fe5e69176313f7c8f0-universal.dmg
is at https://archive.psg.com/emacs25-nightly.trace

>> 13" mid 2014 macbook pro
>> 10.11.6 (15G31)
>> XQuartz 2.7.9 (xorg-server 1.17.4)
>> Emacs-24.5-1-universal.dmg
>>
>> started either by
>>   /Applications/Emacs.app/Contents/MacOS/Emacs at bash cli within an   xterm
>>   or by double clicking the icon
>>
>> with or without ~/.emacs
>>
>> immediately after start, hit C-h to try to find the value of
>> window-system.  crash.
>>
>> what macos gives me for the crash is at
>> https://archive.psg.com/emacs24.trace
>>
>> any clues on how to further debug much appreciated.
>>
>> fwiw, this email written in wanderlust under
>>    GNU Emacs 22.3.1 (i386-apple-darwin9.8.0, Carbon Version 1.6.0) of 2010-01-10 on gs674-seijiz.local
>> which i have been running happily for years

Re: Emacs-24.5-1-universal.dmg for macosx dumps on first keystroke

[Alan Third]

On Fri, Aug 26, 2016 at 04:10:03AM +0900, Randy Bush wrote:
> > I'd recommend you try the 25.1-rc2 release as there have been a lot of
> > crashes fixed.
> 
> the crash dump from Emacs-pretest-25.1-rc2-universal.dmg is at
> https://archive.psg.com/emacs25.1-rc2.trace
> 
> the crash dump from
> Emacs-2016-08-25_01-41-34-70cfe9df957bd2f0699f62fe5e69176313f7c8f0-universal.dmg
> is at https://archive.psg.com/emacs25-nightly.trace

OK, thanks for testing. I can’t reproduce it.

Just to double check, all you’re doing is running Emacs and then
hitting C‐h? Is it only C‐h that causes the crash?

If you run it from the command line using ‘-Q’, does it make any
difference? (I’m expecting no, given what you’ve said previously.)

Are you able to build Emacs from source? Ideally you’d run it in a
debugger so we could find out exactly which line is causing the crash,
but I don’t think the pre‐built one will do that.

-- 
Alan Third

Re: Emacs-24.5-1-universal.dmg for macosx dumps on first keystroke

[Randy Bush]

i may have it.  if i load ~/.emacs with

    (setq ns-control-modifier "alt")

i get the crashes

i will work more on this after dealing with too much morning

randy

Re: Emacs-24.5-1-universal.dmg for macosx dumps on first keystroke

[Randy Bush]

ok, for sure it is 

    (setq ns-just-about-anything "foo")

side issue is that i have lost my mapping of macos command key to meta.
and, given the above, i can not even check if, for example,
    (setq ns-command-modifier "meta")
would work.

in 22, it used to load ../site-lisp/site-start.el which was carbon
specific.  this seems passé.  do i care about lack of, for example,
cocoa replacement?

randy

Re: Emacs-24.5-1-universal.dmg for macosx dumps on first keystroke

[Alan Third]

On Fri, Aug 26, 2016 at 12:11:39PM +0900, Randy Bush wrote:
> ok, for sure it is 
> 
>     (setq ns-just-about-anything "foo")

OK, I can reproduce it now:

emacs -Q

(setq ns-control-modifier "alt")

C‐x C‐e C‐a

and it segfaults.

As a quick fix for you, Randy, instead of using "alt", use 'alt. That
works as you’d expect. I still need to work out why this crashes,
though. Presumably something somewhere is looking for a symbol and
really doesn’t like getting a string instead.
-- 
Alan Third

Re: Emacs-24.5-1-universal.dmg for macosx dumps on first keystroke

[Robert Weiner]

[-- Attachment #1: Type: text/plain, Size: 547 bytes --]

On Fri, Aug 26, 2016 at 2:21 PM, Alan Third <alan@idiocy.org> wrote:

> Presumably something somewhere is looking for a symbol and
> really doesn’t like getting a string instead.
>

​Although experienced users will read the documentation for
ns-control-modifier
​ and recognize that the value should​
​
 be a symbol, there is nothing specific that tells less experienced users
that it must be a symbol and not a string.  An explicit addition of such
language to appropriate ns- variables will eliminate any question.

Bob

[-- Attachment #2: Type: text/html, Size: 1416 bytes --]

Re: Emacs-24.5-1-universal.dmg for macosx dumps on first keystroke

[Alan Third]

On 26 August 2016 at 19:21, Alan Third <alan@idiocy.org> wrote:
> As a quick fix for you, Randy, instead of using "alt", use 'alt. That
> works as you’d expect. I still need to work out why this crashes,
> though. Presumably something somewhere is looking for a symbol and
> really doesn’t like getting a string instead.


I've had a look about and am currently none the wiser. Is there a standard
way of handling this sort of thing in C code? Is it just a case of putting
some sort of SYMBOLP check in before the variable is used or should we
be able to handle a string instead of a symbol?

-- 
Alan Third

Re: Emacs-24.5-1-universal.dmg for macosx dumps on first keystroke

[Randy Bush]

>> As a quick fix for you, Randy, instead of using "alt", use 'alt. That
>> works as you’d expect. I still need to work out why this crashes,
>> though. Presumably something somewhere is looking for a symbol and
>> really doesn’t like getting a string instead.
> 
> I've had a look about and am currently none the wiser. Is there a
> standard way of handling this sort of thing in C code? Is it just a
> case of putting some sort of SYMBOLP check in before the variable is
> used or should we be able to handle a string instead of a symbol?

i concluded that, although a string may have worked in 1948, i should
use `meta in the modern age

randy

Re: Emacs-24.5-1-universal.dmg for macosx dumps on first keystroke

[Anders Lindgren]

[-- Attachment #1: Type: text/plain, Size: 736 bytes --]

>
> On 26 August 2016 at 19:21, Alan Third <alan@idiocy.org> wrote:
> I've had a look about and am currently none the wiser. Is there a standard
> way of handling this sort of thing in C code? Is it just a case of putting
> some sort of SYMBOLP check in before the variable is used or should we
> be able to handle a string instead of a symbol?
>

I would say that we should ensure that Emacs doesn't crash, but it should
ignore all non-symbol values, like most other configuration variables do.

Allowing strings would make the code more complex and it would make testing
harder etc, without giving us any advantages.

The important thing, however, is that the documentation clearly states that
a symbol should be used.

    -- Anders

[-- Attachment #2: Type: text/html, Size: 1129 bytes --]

Re: Emacs-24.5-1-universal.dmg for macosx dumps on first keystroke

[Stefan Monnier]

> I've had a look about and am currently none the wiser. Is there a standard
> way of handling this sort of thing in C code? Is it just a case of putting
> some sort of SYMBOLP check in before the variable is used or should we
> be able to handle a string instead of a symbol?

Usually we use things like CHECK_SYMBOL (foo), but depending on the code
signaling an error might not be a good option, in which case a SYMBOLP
check is indeed the better way.

When set via Custom the user should not be able to set the var to
a string.


        Stefan

Re: Emacs-24.5-1-universal.dmg for macosx dumps on first keystroke

[Eli Zaretskii]

> From: Alan Third <alan@idiocy.org>
> Date: Mon, 29 Aug 2016 10:28:38 +0100
> Cc: Emacs-Devel devel <emacs-devel@gnu.org>
> 
> On 26 August 2016 at 19:21, Alan Third <alan@idiocy.org> wrote:
> > As a quick fix for you, Randy, instead of using "alt", use 'alt. That
> > works as you’d expect. I still need to work out why this crashes,
> > though. Presumably something somewhere is looking for a symbol and
> > really doesn’t like getting a string instead.
> 
> I've had a look about and am currently none the wiser. Is there a standard
> way of handling this sort of thing in C code? Is it just a case of putting
> some sort of SYMBOLP check in before the variable is used or should we
> be able to handle a string instead of a symbol?

This is not NS specific.  The following recipe will crash Emacs on any
platform:

  (let ((where-is-preferred-modifier "alt"))
    (where-is 'goto-char))

I think I've fixed this on the master branch, by outright rejecting
non-symbol values of modifiers in parse_solitary_modifier.  I don't
think supporting strings there is required, as this clearly is a
cockpit error.

Thanks.

Re: Emacs-24.5-1-universal.dmg for macosx dumps on first keystroke

[Alan Third]

On Mon, Aug 29, 2016 at 05:32:20PM +0300, Eli Zaretskii wrote:
> This is not NS specific.  The following recipe will crash Emacs on any
> platform:
> 
>   (let ((where-is-preferred-modifier "alt"))
>     (where-is 'goto-char))
> 
> I think I've fixed this on the master branch, by outright rejecting
> non-symbol values of modifiers in parse_solitary_modifier.  I don't
> think supporting strings there is required, as this clearly is a
> cockpit error.

Thanks Eli. You’ve fixed the crash here.
-- 
Alan Third