Re: Encoding customization variable names

[Gavin Smith]

On Mon, May 30, 2022 at 11:48:22AM +0200, Patrice Dumas wrote:
> On Sun, May 29, 2022 at 02:08:42PM +0100, Gavin Smith wrote:
> > On Fri, Mar 4, 2022 at 7:22 AM Patrice Dumas <pertusus@free.fr> wrote:
> > > Also do not hesitate to propose other names for those customization
> > > variables.
> > 
> > Another issue (I can't remember if I emailed you before about this) is
> > the locale settings to run the test suites. You changed it from "C" to
> > "C.UTF-8", but the latter encoding doesn't exist everywhere. I changed
> > it to "en_US.UTF-8" (see ChangeLog entries on 2022-04-14) but this
> > won't exist everywhere either. Perhaps we should find some way of
> > keeping the test suite under the "C" locale, perhaps disabling or
> > conditionalizing tests that require a UTF-8 locale.
> 
> Some tests need this either to have expected error messages, or
> more fundamentaly for input file names decoding.  I can propose
> something similar with 'Need recoded file names' with a dummy set
> on the test command line.

Could some of this be fixed with customization variables added to the
tests?

For example, when I change the test locale from en_US.UTF-8 to C I
get a test failure

D: formatting//diffs/cpp_lines.diff (printed below)
diff -a -u -r ./formatting//res_parser/cpp_lines/cpp_lines.2 
formatting//out_parser/cpp_lines/cpp_lines.2
--- ./formatting//res_parser/cpp_lines/cpp_lines.2      2022-05-08 
12:15:31.299018608 +0100
+++ formatting//out_parser/cpp_lines/cpp_lines.2        2022-05-30 
18:40:08.891160191 +0100
@@ -1,3 +1,3 @@
 g_f:74: @include: could not find file_with_cpp_lines.texi
-accentêd:7: warning: làng is not a valid language code
+accentêd:7: warning: l?ng is not a valid language code
 cpp_lines.texi: warning: must specify a title with a title command or @top
D: formatting//diffs/cpp_lines.diff (printed above)


but the test passes with the following change:

diff --git a/tp/tests/formatting/list-of-tests 
b/tp/tests/formatting/list-of-tests
index 8965ca597c..a9a7df1555 100644
--- a/tp/tests/formatting/list-of-tests
+++ b/tp/tests/formatting/list-of-tests
@@ -12,7 +12,7 @@ documentlanguage_cmdline documentlanguage.texi 
--document-language=fr
 
 # already tested in t/*.t, but here want to have a result with
 # accented characters in error messages
-cpp_lines ../../t/input_files/cpp_lines.texi
+cpp_lines ../../t/input_files/cpp_lines.texi -c MESSAGE_ENCODING=UTF-8
 
 # some command-line arguments when incorrect cause texi2any to die.
 # easily tested by calling directly ./texi2any.pl and checking visually:
diff --git a/tp/tests/run_parser_all.sh b/tp/tests/run_parser_all.sh
index 3054ba010f..e29a89ee1f 100755
--- a/tp/tests/run_parser_all.sh
+++ b/tp/tests/run_parser_all.sh
@@ -126,8 +126,8 @@ post_process_output ()
   fi
 }
 
-LC_ALL=en_US.UTF-8; export LC_ALL
-LANGUAGE=en_US.UTF-8; export LANGUAGE
+LC_ALL=C; export LC_ALL
+LANGUAGE=en; export LANGUAGE
 
 prepended_command=
 #prepended_command=time


I'm going to try to see if some of the other tests can be made to pass
the same way.

> 
> How can the existence of "en_US.UTF-8" be tested?
> 
> -- 
> Pat