Re: [pdf-devel] Text module unit tests failing

[Aleksander Morgado]

Hoola Jose,

>    > And that's what the DEBUG info says right ?
>    > ##
>    >    libgnupdf: ***DEBUG BASE***:base/pdf-text-context.c:138: Invalid 
> locale
>    >    info detected! 'C'
>    > ###
> 
>    Yes, that's the normal behavior (getting 'C') when you don't call
>    setlocale(LC_ALL,"") (or if your locale environment variables are not
>    set) before calling gl_locale_name, but we are doing it in
>    pdf_text_context_init():
> 
> Do that means that the initialization of the text module will fail if
> LC_ALL= C or LC_ALL = POSIX?

We need to know which is the correct LANG and COUNTRY of the user in the
text module, as a PDF string can contain that information embedded. What
we could do is just assume it to be "en_US" if none available.

We also need to know the 'host encoding' of the user from the locale. If
none available, we could assume it to be UTF-8.

But I don't usually like to make assumptions. It's probably better to
return an error code so that the user knows that the locale is not
correctly set.

> 
> To have such locale configured is quite common. We should be able to
> communicate the user what the problem is, so would be good to add a
> new status return value to 'pdf_text_init' like
> 'PDF_E_BADLOCALE'. What do you think?

Sure, there is no problem to add that, but anyway it will break the unit
tests when LC_ALL=C.

Do you want me to add a sporadic flyspray task for this?

Cheers,
-Aleksander