[PATCH 03 of 10 py3] obsolete: use bytestr() instead of str() so the node is bytes on py3

[Yuya Nishihara]

On Wed, 2 Aug 2017 11:37:45 -0400, Augie Fackler wrote:
> 
> > On Aug 2, 2017, at 11:11, Yuya Nishihara <yuya at tcha.org> wrote:
> > 
> > On Tue, 01 Aug 2017 16:34:31 -0400, Augie Fackler wrote:
> >> # HG changeset patch
> >> # User Augie Fackler <augie at google.com>
> >> # Date 1500907059 14400
> >> #      Mon Jul 24 10:37:39 2017 -0400
> >> # Node ID f50da59f7977c7d796850b95ac7c80a4fd92bd18
> >> # Parent  8b22554199879f80a13ddee556bc540ddcb65a94
> >> obsolete: use bytestr() instead of str() so the node is bytes on py3
> >> 
> >> I'm not sure this is right, since this should either be bytes or str
> >> to match what's going on in the revlog layer.
> > 
> > Sounds like a bug of revlog. 'prec' should be a binary node id, which can't
> > be a unicode string.
> 
> I think in some cases prec coming in as something that __str__ or __repr__s down to a valid node id, which is why the str() is here. It definitely breaks if you just remove the str().

Ah, got it. Perhaps it's a changectx. In which case, I think bytes() should
work. If prec doesn't implement __bytes__(), it's probably wrong.

> >> -        marker = (str(prec), tuple(succs), int(flag), metadata, date, parents)
> >> +        prec = bytes(pycompat.bytestr(prec))
> >> +        marker = (prec, tuple(succs), int(flag), metadata, date, parents)
>