Change Stream type to io.Reader

[aol-nnov]

Wild guesses and musings on #358

@mholt please find my changes concerned the ArchiveFS Stream type.

At first sight, it should not do any harm as zip archiver has type assertion inside and other types do not need a SectionReader (and your zip tests still pass :) )

[mholt]

Thanks for this!

I think this LGTM, but I want @tgulacsi to double-check since I know he was using the SectionReader as of #307, and although I imagine this change should be fine, I want to see what he says first 👍

[tgulacsi]

AFAIU this will advance the underlying Reader on Stat and ReadDir. I don't know whether this will cause problems with a .tar gz or not, just noticed it.

[mholt]

@tgulacsi Thanks, although, I'm not sure I understand why:

this will advance the underlying Reader on Stat and ReadDir.

If the value of the Stream field (type io.Reader) is actually a *SectionReader, wouldn't the behavior remain the same? Why would anything be different? Edit: Oh, or are you talking about the other 3 changes that remove the SectionReader?

[tgulacsi]

Yes, those changes use the Reader, which will advance if Read is called on them. (Not, if a new asectionReader is created.)

[mholt]

Understood. Ok, that makes sense.

@aol-nnov We may need to preserve the SectionReader. I believe calling Read on the underlying reader otherwise is significant.

[aol-nnov]

Okay, @mholt, limitations understood.

[mholt]

Thanks for letting me see the change anyway! :)