Comment 9 for bug 1121254

Yeah, all the current complex flow is stupid and prone to errors.
I don't remember why we are trying to read tags and convert at the same time, but this is way too complex.
I think this was used to remove tags from memory while we were converting, to decrease memory usage. But it's not worth it.

I wanted to fix all this when rewriting soundconverter 3.0 from scratch, but it seems not so hard to fix in fact.
Just read tags if needed, and them convert.

This is exactly what you are proposing BTW. I still have to figure what happens when we have a codec installation...