1652893 - `moz-phab reorg` wants to abandon a closed, landed revision in a stack when the stack is modified after the revision’s initial creation

Status: RESOLVED WONTFIX

(Conduit :: moz-phab, defect)

Description

[Bianca Danforth [:bdanforth]]

I recently encountered a situation where moz-phab reorg wanted to abandon a closed, landed revision.

STR:

1. Create a new phabricator revision, Revision A
   1.1. For me, this was Bug 1638587 (D76073)
2. Create another new phabricator revision, Revision B
   2.1. For me, this was Bug 1643163 (D78591)
3. Create another new phabricator revision, Revision C
   3.1. For me, this was Bug 1612258 (D81377)
4. Rebase the patch for Revision A on top of Revision C, so that now, Revision C is the parent to Revision A
5. moz-phab reorg
   5.1. This correctly identifies the new relationship
6. Land Revision C in mozilla-central
7. Locally pull mozilla-central and hg prune the patch for Revision C.
8. Rebase the patch for Revision A onto central
9. Rebase the patch for Revision B on top of the patch for Revision A
10. moz-phab reorg

Actual results:
moz-phab reorg wants to abandon the closed, landed Revision C (D81377).

    bdanforth ~/src/mozilla-unified $ moz-phab reorg
    Reorganisation based on 2 commits:
    Stack will be reorganised:
     * D81377 will be abandoned
     * D78591 will depend on D76073
    Perform reorganisation (YES/No)?

Expected results:
moz-phab reorg does not want to abandon the closed, landed Revision C (D81377).

I ended up not using moz-phab reorg in this case. The workaround for me was to go into the Phabricator UI for D78591 and update its relationship to D76073 there.

Comment 1

[Piotr Zalewa [:zalun]]

moz-phab reorg role is to update the Phabricator relationship to reflect the local one.
You've changed the local commit stack by removing Revision C. moz-phab reorg wants to do the same on Phabricator to be in sync.
I think it's the right idea to modify this relationship in Phabricator UI.

Please reopen if you think it's an important feature.

Comment 2

[Bianca Danforth [:bdanforth]]

IIUC, moz-phab reorg changes affect the stack in Phabricator, and if a revision is abandoned in Phabricator, it will be hidden by default in Bugzilla. Then if someone later on went to look at that bug in Bugzilla, the published patch would not be readily visible to them in the Bugzilla UI. Is this what would happen? If so, that definitely seems like a bug to me.

Comment 3

[Piotr Zalewa [:zalun]]

In bug 1654837 we're implementing a --keep option which would prevent from abandoning revisions.