GNU bug report logs - #26837
Wrong file in "autoloading failed to define" error

Previous Next

Full log

Message #32 received at 26837 <at> debbugs.gnu.org (full text, mbox):

From: Alex <agrambot <at> gmail.com>
To: Glenn Morris <rgm <at> gnu.org>
Cc: 26837 <at> debbugs.gnu.org, eliz <at> gnu.org
Subject: Re: bug#26837: Wrong file in "autoloading failed to define" error
Date: Thu, 18 May 2017 14:11:00 -0600

Glenn Morris <rgm <at> gnu.org> writes:

> Glenn Morris wrote:
>
> I think it was the lexical-binding thing. Following seems to work:
>
> --- i/src/lread.c
> +++ w/src/lread.c
> @@ -1885,7 +1885,7 @@ PREDICATE can also be an integer to pass to the faccessat(2) function,
>        /* On the first cycle, we can easily test here
>  	 whether we are reading the whole buffer.  */
>        if (b && first_sexp)
> -	whole_buffer = (PT == BEG && ZV == Z);
> +	whole_buffer = (BUF_PT (b) == BUF_BEG (b) && BUF_ZV (b) == BUF_Z (b));
>  
>        instream = stream;
>      read_next:
> @@ -2008,6 +2008,7 @@ BUFFER is the buffer to evaluate (nil means use current buffer),
>    record_unwind_protect (save_excursion_restore, save_excursion_save ());
>    BUF_TEMP_SET_PT (XBUFFER (buf), BUF_BEGV (XBUFFER (buf)));
>    specbind (Qlexical_binding, lisp_file_lexically_bound_p (buf) ? Qt : Qnil);
> +  BUF_TEMP_SET_PT (XBUFFER (buf), BUF_BEGV (XBUFFER (buf)));
>    readevalloop (buf, 0, filename,
>  		!NILP (printflag), unibyte, Qnil, Qnil, Qnil);
>    unbind_to (count, Qnil);

That seems to work for me as well. Should it be documented that
reevaluations of a file update the position of the file in load-history?
I assumed this was already the case when making my autoload patch
because of the "history" in load-history, but documentation would help
to make that clear.

PS: I realized that I was a bit careless in using SDATA without checking
that the argument is necessarily a string. I don't think it matters much
if your above patch is applied, but it's probably not a good idea to
leave it in, just in case there's a bug somewhere else in the future.

What do you think about the following diff? This removes the dependency
on load-history altogether:

--- a/src/eval.c
+++ b/src/eval.c
@@ -2021,7 +2021,10 @@ it defines a macro.  */)
 
       if (!NILP (Fequal (fun, fundef)))
        error ("Autoloading file %s failed to define function %s",
-              SDATA (Fcar (Fcar (Vload_history))),
+              SDATA (Flocate_file_internal (Fcar (Fcdr (fundef)),
+                                            Vload_path,
+                                            Vload_suffixes,
+                                            Qnil)),
               SDATA (SYMBOL_NAME (funname)));
       else
        return fun;

Previous Next

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