Re: [cp-patches] Re: Patch: FYI: fix http Headers oddity

[Bryce McKinlay <mckinlay at redhat dot com>]

Michael Koch wrote:

> > > > Does anyone know why we use the custom "LineInputStream" class instead of, say, BufferedReader here? At one point BufferedReader.readLine() was buggy and could "read too 
> > > > much" and block before returning a line, but I believe we have fixed that. Code re-use would help to avoid bugs like this ;-)
> > > >  
> > > >        
> > > Erm, because LineInputStream is an InputStream dealing with bytes, and
> > > BufferedReader is a Reader dealing with chars.
> > >
> > >      
> > Why is this distinction important? We are converting the result to chars (well, a String) anyway.
> >    
>
> LineInputStream reads one byte and transforms it into an ascii character.
> BufferedReader reads a char (2 bytes) as character.
>  

As I said, this isn't really a big deal because we must convert to chars 
anyway, since we're returning a String. Using the BufferedReader, the 
conversion just happens at a different time.

However, Tromey did point out a good reason on IRC earlier today: 
BufferedReader.readLine() would read beyond the end of the headers on 
the underlying stream. Since the underlying stream may really be bytes 
and not characters, we need a mechanism to reset the stream back to the 
first byte after the headers when it reads past the last EOL.

Bryce