Re: lib/58319: usr.bin/mklocale/Makefile misdefines NBCHAR_MAX

[Collin Funk <collin.funk1%gmail.com@localhost>]

The following reply was made to PR lib/58319; it has been noted by GNATS.

From: Collin Funk <collin.funk1%gmail.com@localhost>
To: Taylor R Campbell <riastradh%NetBSD.org@localhost>
Cc: gnats-bugs%NetBSD.org@localhost,  netbsd-bugs%NetBSD.org@localhost
Subject: Re: lib/58319: usr.bin/mklocale/Makefile misdefines NBCHAR_MAX
Date: Sat, 08 Jun 2024 10:13:45 -0700

 Taylor R Campbell <riastradh%NetBSD.org@localhost> writes:
 
 >> Thanks for testing that. Sorry about the uncleaned directories. What
 >> shell were you using? I'm not too familiar with the shell test framework
 >> written for Gnulib so I might have to forward it to someone else.
 >
 > My interactive shell is zsh, but I doubt whether that's relevant
 > because the scripts look like they are sensibly written to use
 > #!/bin/sh.
 
 You're right. That was a silly question on my part. I fixed the
 -p.<random-chars> issue earlier today [1] [2]. It was a difference
 between GNU Coreutils and NetBSD (and likely others) implementation of
 mktemp with the -t option. NetBSD expects an argument hence the -p
 directory. The tests run since two directories are created but only the
 second gets removed.
 
 > I also just noticed that running `./gnulib-tool --create-testdir ...'
 > had left a directory $TMPDIR/gnulib-python-cache-riastradh lying
 > around.
 
 That directory is intentional. Historically, gnulib-tool has been a
 shell script. Recently the Python implementation has reached a usable
 state and is much faster [3]. Our method of testing involved running
 both the shell and Python version in a copied directory and then diffing
 the output [4]. So "passing" means both versions create the same
 configure.ac, Makefiles, and produce the same standard output.
 
 Since gnulib is often in a subdirectory of another project, when the
 Python version creates a __pycache__ directory and writes the bytecode
 it will cause the test to "fail". For later versions of Python we can
 move the cache to /tmp and for older versions it will be ignored.
 
 TLDR: That directory only gets created once and can be reused, unlike
 the -p.<random-chars> spam.
 
 Thanks for the help!
 
 Collin
 
 [1] https://lists.gnu.org/archive/html/bug-gnulib/2024-06/msg00090.html
 [2] https://git.savannah.gnu.org/cgit/gnulib.git/commit/?id=c07e16b04eca6e678977bc4f7a5358bf98805431
 [3] https://lists.gnu.org/archive/html/info-gnu/2024-04/msg00003.html
 [4] https://git.savannah.gnu.org/cgit/gnulib.git/tree/gnulib-tool?id=12d6338260b7dc041b93bc71351788d7eeb81dd0#n157