[PATCH] revlog: allow custom nodeids in subclasses
Wojciech Lopata
lopek at fb.com
Thu Aug 15 18:29:36 CDT 2013
> From: Matt Mackall [mailto:mpm at selenic.com]
> Sent: Thursday, August 15, 2013 1:29 PM
> To: Wojciech Lopata
> Cc: mercurial-devel at selenic.com
> Subject: Re: [PATCH] revlog: allow custom nodeids in subclasses
>
> On Wed, 2013-08-14 at 09:53 -0700, Wojciech Lopata wrote:
> > # HG changeset patch
> > # User Wojciech Lopata <lopek at fb.com>
> > # Date 1376497815 25200
> > # Wed Aug 14 09:30:15 2013 -0700
> > # Node ID 1e1e57b10f2a1927781408b8c01ab84dc84e8d3f
> > # Parent df54cc67b91bea97c4fc299241a67aeecf099d00
> > revlog: allow custom nodeids in subclasses
> >
> > Modify revlog class to allow its subclasses to use custom method of
> computing
> > nodeids. In particular this change is necesary to implement manifest
> > compression.
> >
> > diff --git a/mercurial/revlog.py b/mercurial/revlog.py
> > old mode 100644
> > new mode 100755
> > --- a/mercurial/revlog.py
> > +++ b/mercurial/revlog.py
> > @@ -943,11 +943,15 @@
> >
> > def _checkhash(self, text, node, rev):
> > p1, p2 = self.parents(node)
> > - if node != hash(text, p1, p2):
> > + if not self._ishashcorrect(text, p1, p2, node):
> > raise RevlogError(_("integrity check failed on %s:%d")
> > % (self.indexfile, rev))
>
> Let's move the raise inside the method. Then the method can be renamed
> to checkhash, with no return code.
>
> > - def addrevision(self, text, transaction, link, p1, p2, cachedelta=None):
> > + def addrevision(self, text, transaction, link, p1, p2, cachedelta=None,
> > + node=None):
>
> Separate patch, please.
>
> > + if not self._ishashcorrect(btext[0], p1, p2, node):
> > raise RevlogError(_("consistency error in delta"))
>
> You'll note this message is less informative than the earlier one.
Earlier message, that contains filename and rev, doesn't fit this situation - here we assert that nodeid and delta passed to _addrevision match (there is no concept of indexfile or rev yet). So I can't extract one 'checkhash' that would provide accurate error message in both situations. Unless I pass error message as argument to checkhash, what somewhat ruins idea of atomic oracle of hash correctness. Please advise.
>
> --
> Mathematics is the supreme nostalgia of our time.
>
More information about the Mercurial-devel
mailing list