GNU bug report logs - #43280
28.0.50; [native-comp] Symbol's function definition is void

Previous Next

Full log

Message #23 received at 43280 <at> debbugs.gnu.org (full text, mbox):

From: Andrea Corallo <akrl <at> sdf.org>
To: Rob Moss <rob.rgm <at> gmail.com>
Cc: 43280 <at> debbugs.gnu.org
Subject: Re: bug#43280: 28.0.50;
 [native-comp] Symbol's function definition is void
Date: Sat, 12 Sep 2020 17:12:23 +0000

Rob Moss <rob.rgm <at> gmail.com> writes:

> On Sat, 12 Sep 2020 at 00:44, Andrea Corallo <akrl <at> sdf.org> wrote:
>>
>> Rob Moss <rob.rgm <at> gmail.com> writes:
>>
>>
>> > Hi Andrea,
>> >
>> > I just tried it with 15acd27d1c0de8b56bab61daa0a8fcd4fef0fdc4 and
>> > experienced the same issue.
>> >
>> > All the best,
>> > Rob
>>
>> Hi Rob,
>>
>> I had some trouble reproducing with the reproduced as I got a lots of
>> errors `Invalid read syntax: "Invalid byte-code object"` byte-compiling
>> use-package.  This both on master and feature/native-comp.
>
> Hi Andrea,
>
> I'm sorry, I don't really know how to make a minimal working example
> that downloads and installs packages. Do you know of any examples that
> I could learn from?

Hi Rob,

tipically what helps is if the reproducer manage to reproduce the issue
from a scratch Emacs.

IOW it should be runnable with something like:

HOME=~/a_tmp_emacs_dir .../emacs -Q -l reproducer.el

>> Anyway I managed to reproduce the issue with elpher installing it
>> manually.
>>
>> I believe the issue is not compiler related, elpher is using `url-port'
>> as a generalized variable (line 1326) but ATM I do not see this defined
>> anywhere in the Emacs codebase (maybe something was changed?).
>
> I'd never heard of a generalized variable, so I've read some of the
> relevant sections of the Emacs manual. I can't tell whether the line
> in question:
>
>     (setf (url-port address) (url-port current-address))
>
> should behave "correctly" (whatever that means). But elpher requires
> 'url-util, which in turns requires 'url-parse. So I would have
> expected `url-port' to be a valid function at this point, because it's
> defined in lisp/url/url-parse.el. I'm particularly confused because I
> only see the warning once elpher has been compiled to native code. So
> there does seem to be some difference in behaviour between the byte
> code and native code, at least on my computer.

Uh I was wrong apologies! :/

The gv setter for `url-port' is defined and is meant to be using
`url-portspec' that is the setter defined by the structure definition
`url' at the top of the url-parse.el.

I see the different behaviour from master to native-comp.  I think this
requires some more investigation on my side.

Thanks for the report!

  Andrea

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.