[ccan] CCAN: code upload with name ogg_to_pcm.tar.gz

[Stephen Cameron]

--- On Wed, 4/8/09, Rusty Russell <rusty at rustcorp.com.au> wrote:
[...]
> Stephen Cameron wrote:
> > 
> > That wasn't gamesmanship, that was me trying to figure out how this ccan
> > thing works.
> 
> Yes, but it was a nice topical point (we'd been discussing
> this recently
> on IRC).  And seeing the games people play lets us
> improve ccanlint :)

Ah, I see.

Yeah, ccanlint is a bit rough around the edges.

I had a comment in _info.c that was "Malformed" in
that it was a completely blank line (no leading ' *')
and I think ccanlint said:

26: Malformed

or something that cryptic.  Had to dig around in the ccanlint code to figure out what it was talking about -- at the point in the code I found it spitting out that message, it didn't seem to have knowledge of what file it was parsing, so couldn't print out the file name easily.

> 
> > Got any bright ideas how to test whether an ogg gets
> decoded correctly?
> > 
> > I thought about embedding ogg and corresponding PCM
> data in static character
> > arrays, writting the ogg data out to a file in /tmp,
> decoding that to memory
> > and memcmp vs. the previously stored PCM data ... but
> if libvorbisfile
> > tweaks their algorithm, that will break, so that's not
> really cool.
> 
> I would go for a sloppy comparison.  Probably using a
> generated sound wave
> sample; you can just store the ogg in the test dir.
> 
> Perfect, no, but I'd expect that the average value to be
> correct,
> and the sample error to be limited (insert some handwave
> stats here).
> 
> This would catch any gross conversion errors, at least.

I'll see what I can come up with.

BTW, how do you prefer changes submitted... patches?  Or just upload a whole new tarball?

-- steve