[PATCH 1 of 4] match: make _fileroots a @propertycache and rename it to _fileset

[Martin von Zweigbergk]

On Fri, May 19, 2017 at 9:07 AM, Yuya Nishihara <yuya at tcha.org> wrote:
> On Fri, 19 May 2017 08:37:06 -0700, Martin von Zweigbergk wrote:
>> On Fri, May 19, 2017 at 7:32 AM, Yuya Nishihara <yuya at tcha.org> wrote:
>> > On Thu, 18 May 2017 11:03:43 -0700, Martin von Zweigbergk via Mercurial-devel wrote:
>> >> # HG changeset patch
>> >> # User Martin von Zweigbergk <martinvonz at google.com>
>> >> # Date 1495123477 25200
>> >> #      Thu May 18 09:04:37 2017 -0700
>> >> # Node ID 48275b23f01283fdf1ae6e630c2c63660528edfc
>> >> # Parent  0d6b3572ad924103128bb9cd296000fc6fd821ef
>> >> match: make _fileroots a @propertycache and rename it to _fileset
>> >
>> > Looks good overall. Queued, thanks.
>> >
>> >> @@ -399,7 +402,6 @@
>> >>                  return matcher.visitdir(self._path)
>> >>              return matcher.visitdir(self._path + "/" + dir)
>> >>          self.visitdir = visitdir
>> >> -        self._fileroots = set(self._files)
>> >
>> > This change should be fine as self._fileset is never be cached in __init__().
>>
>> Even if it were, this would overwrite it (which is what this was meant
>> to do), no?
>
> Doh, I thought subdirmatcher would call match.__init__() before doing the

It will eventually call into its base class's __init__() once I'm done
(although that will be a "basematcher" class, not the "match class
itself)).

> hack, (and if match.__init__() accessed to _fileset, set(_files) != _fileset.)
> But I was wrong. :)

There's a first time for everything :-) (I say that with admiration,
no sarcasm.)