On 11/02/2016 03:45 AM, Jun Wu wrote:
Excerpts from Pierre-Yves David's message of 2016-11-02 03:35:34 +0100:
On 11/02/2016 03:15 AM, Jun Wu wrote:
Excerpts from Pierre-Yves David's message of 2016-11-02 02:53:52 +0100:
I would like something a bit more tangible than "I believe" for example
the changeset that made it a method without cleaning the argument. It
might contains data about why it did not cleaned them up.
See 087603b50889. It does not contain data why it did not cleaned them up.
The commit message does not says anything, but the last hunk probably
have the answer. In the last hunk the path the path, filelog and
filenode are not taken from "self" because we are adjusting link for
parent filelog we are creating. It seems like the code evolve and such
operation do not happen anymore (probably because the parent is lazily
responsible to adjust its likerev if needed.
Can you double check that all trace of this is now gone? If we, I'm
happy to take a signature simplification :-)
I have done this when writing the patch. This is the second check:
Searching all "_adjustlinkrev" from the repo (at 264f00):
quark % ack _adjustlinkrev
mercurial/context.py
683: return self._adjustlinkrev(self._path, self._filelog, <<<
811: def _adjustlinkrev(self, path, filelog, fnode, srcrev,
inclusive=False):
874: return self._adjustlinkrev(self._path, self._filelog, self._filenode,
<<<
878: """create parent filectx keeping ancestry info for
_adjustlinkrev()"""
884: # This lets us later use _adjustlinkrev to get a correct link.
955: # adjustment. Otherwise, p._adjustlinkrev() would walk
changelog
We only have 2 callers of _adjustlinkrev, both using self._path, _filelog
and _filenode.
If patch 4 is dropped, this one will cause conflict. I will fix that later.
I've fixed the patch in flight and pushed it. Thanks.
(only patch5 is pushed)
--
Pierre-Yves David
_______________________________________________
Mercurial-devel mailing list
Mercurial-devel@mercurial-scm.org
https://www.mercurial-scm.org/mailman/listinfo/mercurial-devel
