[PATCH 3 of 4 STABLE] manifest: make manifestctx store the repo

[Durham Goode]

On 10/22/16 1:59 AM, Yuya Nishihara wrote:
> On Tue, 18 Oct 2016 17:50:16 -0700, Durham Goode wrote:
>> # HG changeset patch
>> # User Durham Goode <durham at fb.com>
>> # Date 1476837866 25200
>> #      Tue Oct 18 17:44:26 2016 -0700
>> # Branch stable
>> # Node ID 3efece5c59853908d65de53635488361dbf20c49
>> # Parent  ed607426a3ff4deda8c7f2de8b86d5b6ca976d67
>> manifest: make manifestctx store the repo
>>
>> The old manifestctx stored a reference to the revlog. If the inmemory revlog
>> became invalid, the ctx now held an old copy and would be incorrect. To fix
>> this, we need the ctx to go through the manifestlog for each access.
>>
>> This is the same pattern that changectx already uses (it stores the repo, and
>> accesses commit data through self._repo.changelog).
>>
>> diff --git a/mercurial/manifest.py b/mercurial/manifest.py
>> --- a/mercurial/manifest.py
>> +++ b/mercurial/manifest.py
>> @@ -1276,7 +1276,7 @@ class manifestlog(object):
>>           if self._treeinmem:
>>               m = treemanifestctx(self._revlog, '', node)
>>           else:
>> -            m = manifestctx(self._revlog, node)
>> +            m = manifestctx(self._repo, node)
>>           if node != revlog.nullid:
>>               self._mancache[node] = m
> This will create a reference cycle, but I don't know if the situation gets
> better or worse after this patch.
>
>    repo -> manifestlog -> _mancache -> manifestctx -> repo
>
> Is _mancache valid after the manifestlog is recreated?
_mancache is not really valid after manifestlog is recreated, since it 
may contain manifest entries that no longer exist in the file. Even if 
it didn't contain bad entries, the manifestctx itself needs up-to-date 
access to the manifest revlog, which is only available through the repo 
object (since the repo object's property is what handles manifest revlog 
cache checking).