Bug#784639: lintian: Reports on lintian.debian.org contain non-UTF8 encoding

[Niels Thykier <niels@thykier.net>]

Control: tags -1 confirmed
Control: notfound -1 2.5.30+deb8u4

On 2015-05-07 13:47, Andreas Ronnquist wrote:
> Package: lintian
> Version: 2.5.30+deb8u4
> Severity: minor
> 
> Dear Maintainer,
> 
> Being cursed with a surname containing non-ASCII characters, I notice
> that the reports on lintian.debian.org [1], encodes the 'ö' in my
> surname wrongly, which causes browsers set on UTF8 to show it wrongly.
> I have tested with Iceweasel from stable, and (non-debian-packaged)
> Midori.
> 
> [1] https://lintian.debian.org/maintainer/gusnan@debian.org.html
> [...]
> 

Hi,

I have localised this to YAML::Any being useless when it comes to
encoding.  It turns out that YAML::XS (my local implementation) and
whatever YAML::Any picks on lindsay.d.o wildly disagrees on how to deal
with encoding:

 * YAML::XS croaks if given "decoded" text, but automatically converts
   byte streams into UTF-8 encoded YAML data structures
 * YAML::Something produces mojibake if given a "non-decoded" text (i.e.
   a byte stream) and only produces the correct result if given UTF-8
   text.

In conclusion - YAML::Any is /not/ a consistent interface of
interchangeable implementations, which I originally thought it was.  I
have asked DSA to install YAML::XS on lindsay.d.o and I intend to
replace YAML::Any with YAML::XS  in the code once that is done.

I have removed the found version as it is (strictly) speaking an
artefact of the unreleased version, which is currently running on
lindsay.d.o.

Thanks,
~Niels