[PATCH] localrepo: actually invalidate dirstate on invalidatedirstate()

[Mateusz Kwapich]

So invalidate is by design not-always invalidating and it’s relying on filecache to actually invalidate dirstate
when needed? Why do we even need the invalidate in this case if filecache can always properly invalidate dirstate?

Best,
Mateusz

On 8/26/16, 5:02 PM, "Yuya Nishihara" <youjah at gmail.com on behalf of yuya at tcha.org> wrote:

    On Fri, 26 Aug 2016 08:31:09 -0700, Mateusz Kwapich wrote:
    > # HG changeset patch
    > # User Mateusz Kwapich <mitrandir at fb.com>
    > # Date 1472225341 25200
    > #      Fri Aug 26 08:29:01 2016 -0700
    > # Node ID 430e8c2cc229e0fbed231370c36cc2a215ecb30e
    > # Parent  318e2b600b80e4ed3c6f37df46ec7544f60d4c0b
    > localrepo: actually invalidate dirstate on invalidatedirstate()
    > 
    > The old dirstate was still present in the filecache. It was invalidated only
    > because filecache did stat the file on the next access. This can lead to errors
    > in the case when file stat didn't change but the dirstate contents did change
    > (and we know about it and thats why we sometimes call dirstateinvalidate() to
    > refresh it).
    
    (+CC foozy since he's working on fixing timestamp issues)
    
    IIRC, that's by design. invalidate() just forces to compare timestamps to see
    if in-memory cache is valid.