bug#52322: `make --question info` always fails

bug#52322: `make --question info` always fails

[Stefan Kangas]

This command on emacs-28 and master always fails:

    make --question info

Here is a reproducer:

    make info && make --question info || echo "failed"

This breaks the `make-dist` script, where we rely on `make --question
info' returning 0 to continue.  I had to comment out that line to
produce the first pretest.

I have bisected this error to the following commit:

    fddd63f8b854f6bfa91403f69ba694ccb54197bc is the first bad commit
    commit fddd63f8b854f6bfa91403f69ba694ccb54197bc
    Author: Glenn Morris <rgm@gnu.org>
    Date:   Fri Feb 26 19:28:43 2021 -0800

        Distribute the real source for some doc/misc manuals (bug#45143)

        * doc/misc/modus-themes.texi, doc/misc/org.texi:
        Remove generated files from repository.
        * doc/misc/Makefile.in: Add rules for building .texi from .org.
        (ORG_SRC, abs_top_builddir, EMACS, emacs):
        New variables.
        (org_template): New template.
        (orgclean): New phony target.
        * Makefile.in (info): Depend on lisp.
        * lisp/org/ox-texinfo.el (org-texinfo-export-to-texinfo-batch):
        New function.
        * doc/misc/org.org, doc/misc/org-setup.org: New files.
        Import from https://code.orgmode.org d8e8a97a14.

bug#52322: `make --question info` always fails

[Glenn Morris]

ISTM that src/Makefile.in has a problem, in that it wants globals.h to
be newer than gl-stamp, but the rule that generates globals.h ensures
the opposite is always true. So it always tries to rebuild globals.h
(even if it is a no-op).

Working around that, one is left with the issue that compile-clean
always runs, and make can't tell that it can be a no-op.

Perhaps just get rid of the --question test.

bug#52322: `make --question info` always fails

[Stefan Kangas]

Glenn Morris <rgm@gnu.org> writes:

> ISTM that src/Makefile.in has a problem, in that it wants globals.h to
> be newer than gl-stamp, but the rule that generates globals.h ensures
> the opposite is always true. So it always tries to rebuild globals.h
> (even if it is a no-op).
>
> Working around that, one is left with the issue that compile-clean
> always runs, and make can't tell that it can be a no-op.
>
> Perhaps just get rid of the --question test.

Fine by me.  AFAICT, we will run "make info" anyways.

Any other comments?

bug#52322: `make --question info` always fails

[Stefan Kangas]

[-- Attachment #1: Type: text/plain, Size: 375 bytes --]

close 52322 28.1
thanks

Stefan Kangas <stefan@marxist.se> writes:

> Glenn Morris <rgm@gnu.org> writes:
>
>> Perhaps just get rid of the --question test.
>
> Fine by me.  AFAICT, we will run "make info" anyways.
>
> Any other comments?

There have been no further comments within over a week, so I've
installed the attached patch on the emacs-28 branch (commit 9de8eec3bc).

[-- Attachment #2: 0001-make-dist-Don-t-run-make-question-info-.-Bug-52322.patch --]
[-- Type: text/x-diff, Size: 786 bytes --]

From 9de8eec3bc9be2f6e7debf26159c1761c47df909 Mon Sep 17 00:00:00 2001
From: Stefan Kangas <stefan@marxist.se>
Date: Tue, 14 Dec 2021 03:08:37 +0100
Subject: [PATCH] * make-dist: Don't run "make --question info".  (Bug#52322)

---
 make-dist | 7 -------
 1 file changed, 7 deletions(-)

diff --git a/make-dist b/make-dist
index b069130ca6..db7a74b82b 100755
--- a/make-dist
+++ b/make-dist
@@ -299,13 +299,6 @@ tempparent=
     echo "${bogosities}"
   fi
 
-  ## This exits with non-zero status if any .info files need
-  ## rebuilding.
-  if [ -r Makefile ] && [ "$with_info" = "yes" ]; then
-      echo "Checking to see if info files are up-to-date..."
-      make --question info || error=yes
-  fi
-
   ## Is this a release?
   case $version in
       [1-9][0-9].[0-9])
-- 
2.30.2