GNU bug report logs - #56746
29.0.50; Probable bug with native compilation

Previous Next

Full log

Message #20 received at 56746 <at> debbugs.gnu.org (full text, mbox):

From: Andrea Corallo <akrl <at> sdf.org>
To: Kenneth Stuart <kstuart <at> hotmail.co.uk>
Cc: 56746 <at> debbugs.gnu.org, Eli Zaretskii <eliz <at> gnu.org>
Subject: Re: bug#56746: 29.0.50; Probable bug with native compilation
Date: Wed, 27 Jul 2022 08:13:39 +0000

Kenneth Stuart <kstuart <at> hotmail.co.uk> writes:

> Andrea Corallo <akrl <at> sdf.org> writes:
>
>> Eli Zaretskii <eliz <at> gnu.org> writes:
>>
>>>> From: Kenneth Stuart <kstuart <at> hotmail.co.uk>
>>>> Date: Sun, 24 Jul 2022 15:16:42 +0100
>>>> 
>>>> Note that I'm not using 'emacs -Q' as a later version of Org is
>>>> required, however the issue strongly appears to be with emacs native
>>>> compilation as it does not occur from a build without native
>>>> compilation.
>>>> 
>>>> org-mode commit: 127e7fee4959ceb33dba2e442b72189f3c25eaa1
>>>> 
>>>> Steps to reproduce:
>>>> 
>>>> 1. Create an org-mode buffer and insert a clocktable dynamic block with a
>>>> `tstart` option.
>>>> 
>>>> #+BEGIN: clocktable :scope file :maxlevel 2 :tstart "2020-01-01"
>>>> #+END:
>>>> 
>>>> 2. Evaluate the clocktable (C-c C-c on the #+BEGIN line)
>>>> 
>>>> Result: `Invalid function: org-encode-time`
>>>> Expected: No error, clocktable updated successfully
>>>> 
>>>> Notes:
>>>> 
>>>> References:
>>>> https://mail.gnu.org/archive/html/emacs-orgmode/2022-07/msg00738.html
>>>> https://mail.gnu.org/archive/html/emacs-orgmode/2022-07/msg00696.html
>>>> 
>>>> In early April Org introduced a new macro `org-encode-time` which is
>>>> used in the `let` special form of function `org-matcher-time`
>>>> 
>>>> https://git.savannah.gnu.org/cgit/emacs/org-mode.git/tree/lisp/org-macs.el#n1395
>>>> https://git.savannah.gnu.org/cgit/emacs/org-mode.git/tree/lisp/org-macs.el#n1463
>>>> 
>>>> Function `org-matcher-time` is used by the clocktable dynamic block
>>>> options, for example adding a `tstart` option and evaluation the
>>>> clocktable emits the error "Invalid function: org-encode-time".
>>>> 
>>>> If then going to the source for `org-matcher-time` and evaluating the
>>>> defun, the clocktable can then be evaluated without error.
>>>> 
>>>> As previously stated this issue does not occur when not using native
>>>> compilation.
>>>
>>> Andrea, could you please look into this bug report?  It sounds like
>>> 'defmacro' in a conditional code is mis-compiled somehow?
>>
>> I'm currently on holiday so I've only time to check the mails and reply
>> when necessary.  I won't have time to look into this before next week
>> *but*... if it's a macro issue it might be some missing require in the
>> org code, we have seen this kind of issue is the most common.
>>
>> Is this org async compiled?  If so Kenneth have we verifed the emitted
>> warnings?
>>
>> Thanks!
>>
>>   Andrea
>
> During async compile there are a couple of warnings emitted:
>
> In end of data:
> org-macs.el:1463:29: Warning: the function ‘org-encode-time’ is not known to be defined.
> org-macs.el:1361:10: Warning: the function ‘org-time-string-to-seconds’ is not known to be defined.

Nice, those are clear markers of this issue.

> However org-encode-time is earlier defined in org-macs.el (within 'if' statement).

Okay, I can't read the code now but I guess should be easy to
investigate why during async compilation this 'if' is not triggered.

> Also interesting, if I touch org-macs.el and synchronously compile, I see no issue:

This is because compiling synchronously the definition of those macros
is most likely loaded and already present in the environment.

  Andrea

Previous Next

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