[PATCH 2 of 2 v2] acl: support revsets [RFC]

[Augie Fackler]

> On May 11, 2016, at 6:12 AM, Pierre-Yves David <pierre-yves.david at ens-lyon.org> wrote:
> 
>>> Afaict, it does not help w/ bookmarks at all,
>>> as bookmarks aren't applied until after the hook runs.
>>> And I don't think one can roll back a bookmark by that point,
>>> as I think the previous information would be lost.
>> Hm, you might well be right. Sounds like maybe we should consider both
>> patches? Being able to use revsets here seems about right to me feel-wise.
> 
> This series creates the following though flow on my side:
> 
> 1) Why do user need to setup hook by hand that seems like something acl should do by itself

I don’t know what this means.

> 2) Bookmark handling would be better/safer at the transaction level, but we don't have good hook data at that point

I think the problem is not the transaction-level business for bookmarks, but instead the fact that old clients send bookmarks in a separate transaction.

> 3) If we improve hooks used, user will have to adjust config to take advantage of it (because of (1))
> 
> This create a tickling temptation to rework acl to register hooks and improve transaction to track bookmark event before moving forward here.

I understand that temptation, but I think that’s out of scope for what this series is trying to accomplish (and not entirely feasible anyway with old clients potentially in play.)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 801 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <http://www.mercurial-scm.org/pipermail/mercurial-devel/attachments/20160511/cd5db84f/attachment.sig>