Bug#803724: licensecheck still broken

[Nicholas Bamber]

On 15. 11. 15 17:56, Dominique Dumont wrote:
> Hello
>
> Unfortunately, licensecheck is still broken: the encoding detection based on
> `file` command does not work as expected.
>
> For instance, in icedove. the file mozilla/js/src/vtune/jitprofiling.h contains
> windows-1252 character in a comment:
>
> $ file --mime-type --mime-encoding mozilla/js/src/vtune/jitprofiling.h
> mozilla/js/src/vtune/jitprofiling.h: text/plain; charset=unknown-8bit
>
> By default, licensecheck refuses to parse files with unknown encoding.
>
> I'm now convinced that using `file` to detect charset is more trouble than it's
> worth.
>
> Unless someone has a better idea, I'm going to revert most of the modification
> I've done with regard to encoding. licensecheck will go back to:
> - use a regexp to decide whether to parse a file or not
> - don't change encoding of copyright owners.
>
> The latter means that user will have to deal with any non utf8 character
> produced by licensecheck.
>
> All the best
>

Dominique,
	Sorry I have not replied.

	The thing is I don't care how licensecheck copes with difficult cases 
like pdf files or whatever. I know some people do and obviously you need 
to engange with them. all I care about is that when licensecheck is in 
recursive mode it does not miss any files which it would handle 
correctly if asked explicitly. I have had the impression it is prone to 
this, but I have been remiss in not checking this out more throughly. 
Next time I get this impression I will persue it. I almost thought I had 
found this recently but it turned out to be my misunderstanding so for 
now I have no feedback.