bug#81480: 31.0.91; diff-mode-test-* incorrectly assumes "diff -a"

Sean Whitton <[email protected]> Thu, 30 Jul 2026 12:17:28 +0100
Newsgroups gmane.emacs.bugs
Message-ID <[email protected]>
Eli Zaretskii [27/Jul  4:57pm +03] wrote:
>> From: Sean Whitton <[email protected]>
>> Date: Mon, 27 Jul 2026 10:30:28 +0100
>>
>> Paul Eggert [24/Jul  1:55pm -07] wrote:
>> > This follows up on bug#81465 "31.0.91; ‘make check’ failures on Solaris
>> > 10 sparc 32-bit" and I am splitting this issue off into a separate bug
>> > report.
>> >
>> > Some diff-mode-test-* patches fail because they use ‘diff -a’. The -a
>> > option is a GNU extension to POSIX and is not supported by Solaris 10 diff.
>>
>> I don't understand what you mean by saying that they *use* 'diff -a' --
>> if it's -a that turns on the "No newline at end of file" output, then
>> the problem has to be that -a is the implicit default behaviour for GNU
>> diff?
>
> No, AFAICT, the "No newline at end of file" is a separate problem.
>
>> AFAICT we don't explicitly pass -a anywhere.
>
> We do, in smerge-refine-regions (which these tests invoke indirectly,
> see the backtraces from the failing tests).

Thanks, I got confused somehow there.  Let's fix this along the lines
of your patch.

-- 
Sean Whitton