Re: texinfo-6.8.90 pretest

[pertusus]

On Sun, Oct 23, 2022 at 01:42:22PM +0300, Eli Zaretskii wrote:
> On a hunch, I did two things:
> 
>   . created a file tp/tests/input_file_names_recoded_stamp.txt with
>     the text FAILED in it;

Ok.  I'll try to change the generation of input_file_names_recoded_stamp.txt
such that there is no UTF-8 encoded characters in Makefile.am, which
is the actual issue.

>   . made the change below in tp/Texinfo/Translations.pm.
> 
> Then running a single test with
> 
>   $ ./run_parser_all.sh -dir contents contents_at_begin
> 
> succeeded.  Will now try running more tests with these tricks.
> 
> Here's the diff to Translations.pm, it is AFAIU needed because the
> Windows locale names don't follow the Posix patterns.  Instead, they
> are 3-letter names, like "ENU" for en_US, "FRA" for fr_FR, etc.

Ok.  Hopefully it does not interfere with translations, only with file
names encodings.

>  The
> diff below should be obviously conditioned by this being MS-Windows.

Is this locale supposed to always exist on Windows?

I am not sure that this patch is needed, as it seems to me that it
modifies a code that is no longer run on windows with a patch Gavin
applied very recently.

> 
> --- ./tp/Texinfo/Translations.pm~0    2022-10-19 18:30:05.000000000 +0300
> +++ ./tp/Texinfo/Translations.pm      2022-10-23 13:28:19.459500000 +0300
> @@ -96,6 +96,9 @@ sub gdt($$;$$$)
>    if (!$locale) {
>      $locale = POSIX::setlocale(LC_ALL, "en_US")
>    }
> +  if (!$locale) {
> +    $locale = POSIX::setlocale(LC_ALL, "ENU")
> +  }
>    our $locale_command;
>    if (!$locale and !$locale_command) {
>      $locale_command = "locale -a";