Re: [rt.cpan.org #73177] ->show_text does not handle latin1 strings correctly

[Torsten Schoenfeld <kaffeetisch gmx de>]

On 11.12.2011 18:40, Slaven_Rezic via RT wrote:
> Sun Dec 11 12:40:59 2011: Request 73177 was acted upon.
> Transaction: Ticket created by SREZIC
>        Queue: Cairo
>      Subject: ->show_text does not handle latin1 strings correctly
>    Broken in: 1.081
>     Severity: (no value)
>        Owner: Nobody
>   Requestors: SREZIC cpan org
>       Status: new
>  Ticket <URL: https://rt.cpan.org/Ticket/Display.html?id=73177 >
>
>
> The ->show_text method does not handle codepoints > 128 in a perl string
> without utf8 flag correctly. See the attached sample script.
>
> A quick glance in Cairo.xs shows that SvPV is used. Probably SvPVutf8
> should be used instead here to fix the problem. See also "perldoc perlapi".

Yeah, the current implementation basically burdens the user with 
ensuring correct encoding.  As you show in the attached program, that 
basically boils down to utf8::upgrade() if you have latin1-encoded 
strings.  I agree that it would be convenient if Cairo auto-upgraded 
strings if utf8 is expected.  But I worry about backwards-compatibility:

* If someone used utf8::encode() instead of utf8::upgrade(), then 
auto-upgrading will lead to double encoding.

* If someone used utf8 literals without specifying "use utf8", then 
auto-upgrading will also lead to double encoding.

Technically, both these usages are incorrect.  But they work OK with 
current Cairo and it would break if implemented auto-upgrading.

What do you think?  I pushed a branch to the git repo which implements 
the necessary changes: 
<http://git.gnome.org/browse/perl-Cairo/commit/?h=auto-utf8&id=f40ee313502510aedcd2df1a225fa6095717c2d2>. 
 I also CC the mailing list for further opinions.