From: Jakub Narebski <jnareb@gmail.com>
Date: Wed, 06 Oct 2010 23:26:25 GMT
Subject: Re: [PATCHv5 06/17] gitweb/lib - Simple select(FH) based output capture
Message-ID: <201010070126.26242.jnareb@gmail.com>
In-Reply-To: <AANLkTikVumyaJoM3=Nbf9igVheLvMB5tcc5+E+wSkb40@mail.gmail.com>

On Thu, 7 Oct 2010 01:03, Ævar Arnfjörð Bjarmason wrote:
> On Wed, Oct 6, 2010 at 22:01, Jakub Narebski <jnareb@gmail.com> wrote:
> 
> > * The most important issue is that I/O "layers" (PerlIO), like ':utf8'
> >  or ':raw', are *already applied* to the output that is captured.
> >  This means that captured output is *always* in binary (':raw') mode.
> >  In Perl 6 language it means that data returned by capturing engine
> >  is an equivalent of Buf, a collection of bytes, whether Buf or Str
> >  (a colection of logical characters) is printed.
> 
> > +       # note: this does not cover all cases
> > +       binmode select(), ':utf8'
> > +               if ((PerlIO::get_layers($self->{'oldfh'}))[-1] eq 'utf8');

Sidenote: I just realized that we can simply try to replay all but
special layers from 'oldfh'.
 
> I'm not sure but maybe we want to use ":Encoding(UTF-8)" everywhere in
> this series where you've used ":utf8". I.e. use the Encoding UTF-8
> layer instead of the internal utf8 layer.
> 
> It's more strict, see perldoc Encode's "UTF-8 vs. utf8 vs. UTF8". But
> maybe we don't care.

I think for the time being we don't care. 

In the future we could perhaps update gitweb to use ':encoding(UTF-8)'
PerlIO layer rather than ':utf8' layer, and update capturing engine(s)
appropriately.

-- 
Jakub Narebski
Poland