GNU bug report logs - #73748
31.0.50; Problems with company-dabbrev in GNU ELPA package company

Previous Next

To reply to this bug, email your comments to 73748 AT debbugs.gnu.org.

Toggle the display of automated, internal messages from the tracker.

Message #5 received at submit <at> debbugs.gnu.org (full text, mbox):

From: Stephen Berman <stephen.berman <at> gmx.net>
To: bug-gnu-emacs <at> gnu.org
Subject: 31.0.50; Problems with company-dabbrev in GNU ELPA package company
Date: Fri, 11 Oct 2024 13:42:54 +0200

[Message part 1 (text/plain, inline)]

0. Install the GNU ELPA packages company and gnu-elpa.
1. emacs -Q
2. M-x package-initialize and M-x toggle-debug-on-error
3. C-x b a M-x company-mode (switch to buffer "a" in fundamental-mode
   and enable company-mode there)
4. Type three arbitrary letters, e.g. dkf
5. On typing the third letter, the following prompt appears in the echo
   area: "Install package all? (yes or no) ".  Type "no".
==> This raises the error "Abort!" and pops up a *Backtrace* buffer,
whose content is attached.

Looking at the backtrace here is what I think happens (though I couldn't
actually get Edebug to stop execution before the prompt in step 5):
company-dabbrev--fetch call company-dabbrev--search, in which the sexp
`(eq other-buffer-modes 'all)' is evaluated as t ('all being the default
value of company-dabbrev-other-buffers), and this causes
gnu-elpa--perform-autoload (in gnu-elpa-utils.el) to be invoked, which
invokes the install package prompt in step 5 because the symbol `all' is
in gnu-elpa-features.el (from the GNU ELPA package "all").

It would not be unreasonable to lay the blame for this problem on the
"all" package with its namespace-unclean autoloaded command `all'
(naming it something like `all-show' would have avoided the problem).
However, the pcase code in company-dabbrev--fetch actually triggers the
problem, and `M-x eval-buffer' on company-dabbrev.el points to why: that
invocation issues the warning "pcase pattern `all shadowed by previous
pcase pattern".

The attached patch (I don't have a local clone of the company package
repo, so it's just a plain diff) avoids both this warning and the
problematic interaction with gnu-elpa--perform-autoload.  This patch
uses 'all instead of `all, since they are the same for symbols in pcase
patterns, and replaces `t by _, in accordance with (info "(elisp) pcase
Macro").  In addition, the patch fixes a typo in the defcustom
`company-dabbrev-other-buffers' that prevents choosing "Function to
return similar major-modes" from the value menu.


In GNU Emacs 31.0.50 (build 1, x86_64-pc-linux-gnu, GTK+ Version
 3.24.43, cairo version 1.18.2) of 2024-10-05 built on strobelfssd
Repository revision: 4c7a6dc1a0ce8fe7bb603c907efb52d2ecf30858
Repository branch: master
Windowing system distributor 'The X.Org Foundation', version 11.0.12101013
System Description: Linux From Scratch r12.2-17-systemd

Configured using:
 'configure -C 'CFLAGS=-Og -g3' PKG_CONFIG_PATH=/opt/qt6/lib/pkgconfig'

Configured features:
ACL CAIRO DBUS FREETYPE GIF GLIB GMP GNUTLS GPM GSETTINGS HARFBUZZ JPEG
LCMS2 LIBSYSTEMD LIBXML2 MODULES NATIVE_COMP NOTIFY INOTIFY PDUMPER PNG
RSVG SECCOMP SOUND SQLITE3 THREADS TIFF TOOLKIT_SCROLL_BARS TREE_SITTER
WEBP X11 XDBE XIM XINPUT2 XPM GTK3 ZLIB

[company-dabbrev-backtrace (application/octet-stream, attachment)]

[Message part 3 (text/x-patch, attachment)]

Previous Next

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