[Bug 755301] audioconvert: Integer->Float conversion creates values slightly smaller than -1.0

[GStreamer (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=755301

--- Comment #19 from Sebastian Dröge (slomo) <slomo at coaxion.net> ---
(In reply to sebastien.fenet from comment #17)
> Ok, I see. In my opinion, it is better to work with 64 bits float as
> intermediate format.

64 bit floats are much slower than 32 bit integers, especially on embedded
platforms :)

> Anyway, I agree that 16 bits -> 32 bits is lossless but you don't have the
> same problem. 
> [...]

Indeed, good point!

(In reply to sebastien.fenet from comment #18)
> Sorry for multi-posting but I found the answer to my question by myself...
> 
> The initial test I made was Int16 -> Float64.
> 
> I've just tested Int16 -> Float32 and it works as expected!! That is because
> of Sebastian's remark... The dividing factor was wrong but as
> float32(wrong_factor)=float32(good_factor), the result is still OK.

Yes, I guess it's good then. For some reason I thought you observed the problem
with 32 bit floats

> So, Sebastian, I would say that your fix in the code is OK. I was not able
> to find the unit test in log in your previous posts... What is wrong with
> the tests?

Run make elements/audioconvert.check in gst-plugins-base/tests/check. There are
some tests that are expecting the current behaviour, they should just be
updated :)

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.