Re: [PATCH v5] read-cache: do not trust a size change when conversion is active

[Junio C Hamano <gitster@pobox.com>]

Torsten Bögershausen <tboegi@web.de> writes:

> Looking at CE_MATCH_RACY_IS_DIRTY it seems as if the call
> to would_convert_to_git()may be better done in ie_match_stat() ?

It would be a better fit than the patch being discussed, even though
I do not know if that is the best spot.

> Beside that I was able to test it with a real live repo,
> where a clean filter replaces a TAB with 4 spaces: Now
> git status no longer claims the file to be modified
> (and refuses to stash or other operations))

The issue I was raising in the message you are responding to was not
about correctness but about avoiding unnecessary cycles.

If you rip out the lstat()-based optimization altogether and rewrite
the code to always compare the contents, status would not say the
file is modified in the same case, either.  "Now it behaves
differently" is not something we need to boast about.  Having the
code behave the way we want is a mere baseline requirement, and
by v5 we hopefully have passed that point long ago.