phases: avoid a potentially costly dictionary interation in some case

If we retract for the draft phase, there is not non-public item to be retracted
and we can skip this part. This part is was apparently super costly thanks to
Python.

On mozilla-try-2019-02-18, a perf::unbundle call with a 10 000 changesets
bundle gives give use the following timing.

e57d4b868a3e:     4.6 seconds
ac1c75188440:   102.5 seconds
this-changeset:  30.0 seconds

So we recovered about ⅔ of the regression, the next changeset will give us the
rest back.

mercurial/phases.py

@@ -1,1223 +1,1224 b''
 """ Mercurial phases support code
 
     ---
 
     Copyright 2011 Pierre-Yves David <[e-mail]>
                    Logilab SA        <[e-mail]>
                    Augie Fackler     <[e-mail]>
 
     This software may be used and distributed according to the terms
     of the GNU General Public License version 2 or any later version.
 
     ---
 
 This module implements most phase logic in mercurial.
 
 
 Basic Concept
 =============
 
 A 'changeset phase' is an indicator that tells us how a changeset is
 manipulated and communicated. The details of each phase is described
 below, here we describe the properties they have in common.
 
 Like bookmarks, phases are not stored in history and thus are not
 permanent and leave no audit trail.
 
 First, no changeset can be in two phases at once. Phases are ordered,
 so they can be considered from lowest to highest. The default, lowest
 phase is 'public' - this is the normal phase of existing changesets. A
 child changeset can not be in a lower phase than its parents.
 
 These phases share a hierarchy of traits:
 
             immutable shared
     public:     X        X
     draft:               X
     secret:
 
 Local commits are draft by default.
 
 Phase Movement and Exchange
 ===========================
 
 Phase data is exchanged by pushkey on pull and push. Some servers have
 a publish option set, we call such a server a "publishing server".
 Pushing a draft changeset to a publishing server changes the phase to
 public.
 
 A small list of fact/rules define the exchange of phase:
 
 * old client never changes server states
 * pull never changes server states
 * publish and old server changesets are seen as public by client
 * any secret changeset seen in another repository is lowered to at
   least draft
 
 Here is the final table summing up the 49 possible use cases of phase
 exchange:
 
                            server
                   old     publish      non-publish
                  N   X    N   D   P    N   D   P
     old client
     pull
      N           -   X/X  -   X/D X/P  -   X/D X/P
      X           -   X/X  -   X/D X/P  -   X/D X/P
     push
      X           X/X X/X  X/P X/P X/P  X/D X/D X/P
     new client
     pull
      N           -   P/X  -   P/D P/P  -   D/D P/P
      D           -   P/X  -   P/D P/P  -   D/D P/P
      P           -   P/X  -   P/D P/P  -   P/D P/P
     push
      D           P/X P/X  P/P P/P P/P  D/D D/D P/P
      P           P/X P/X  P/P P/P P/P  P/P P/P P/P
 
 Legend:
 
     A/B = final state on client / state on server
 
     * N = new/not present,
     * P = public,
     * D = draft,
     * X = not tracked (i.e., the old client or server has no internal
           way of recording the phase.)
 
     passive = only pushes
 
 
     A cell here can be read like this:
 
     "When a new client pushes a draft changeset (D) to a publishing
     server where it's not present (N), it's marked public on both
     sides (P/P)."
 
 Note: old client behave as a publishing server with draft only content
 - other people see it as public
 - content is pushed as draft
 
 """
 
 
 import heapq
 import struct
 import typing
 import weakref
 
 from typing import (
     Any,
     Callable,
     Dict,
     Iterable,
     List,
     Optional,
     Set,
     Tuple,
 )
 
 from .i18n import _
 from .node import (
     bin,
     hex,
     nullrev,
     short,
     wdirrev,
 )
 from . import (
     error,
     pycompat,
     requirements,
     smartset,
     txnutil,
     util,
 )
 
 Phaseroots = Dict[int, Set[int]]
 PhaseSets = Dict[int, Set[int]]
 
 if typing.TYPE_CHECKING:
     from . import (
         localrepo,
         ui as uimod,
     )
 
     # keeps pyflakes happy
     assert [uimod]
 
     Phasedefaults = List[
         Callable[[localrepo.localrepository, Phaseroots], Phaseroots]
     ]
 
 
 _fphasesentry = struct.Struct(b'>i20s')
 
 # record phase index
 public: int = 0
 draft: int = 1
 secret: int = 2
 archived = 32  # non-continuous for compatibility
 internal = 96  # non-continuous for compatibility
 allphases = (public, draft, secret, archived, internal)
 trackedphases = (draft, secret, archived, internal)
 not_public_phases = trackedphases
 # record phase names
 cmdphasenames = [b'public', b'draft', b'secret']  # known to `hg phase` command
 phasenames = dict(enumerate(cmdphasenames))
 phasenames[archived] = b'archived'
 phasenames[internal] = b'internal'
 # map phase name to phase number
 phasenumber = {name: phase for phase, name in phasenames.items()}
 # like phasenumber, but also include maps for the numeric and binary
 # phase number to the phase number
 phasenumber2 = phasenumber.copy()
 phasenumber2.update({phase: phase for phase in phasenames})
 phasenumber2.update({b'%i' % phase: phase for phase in phasenames})
 # record phase property
 mutablephases = (draft, secret, archived, internal)
 relevant_mutable_phases = (draft, secret)  # could be obsolete or unstable
 remotehiddenphases = (secret, archived, internal)
 localhiddenphases = (internal, archived)
 
 all_internal_phases = tuple(p for p in allphases if p & internal)
 # We do not want any internal content to exit the repository, ever.
 no_bundle_phases = all_internal_phases
 
 
 def supportinternal(repo: "localrepo.localrepository") -> bool:
     """True if the internal phase can be used on a repository"""
     return requirements.INTERNAL_PHASE_REQUIREMENT in repo.requirements
 
 
 def supportarchived(repo: "localrepo.localrepository") -> bool:
     """True if the archived phase can be used on a repository"""
     return requirements.ARCHIVED_PHASE_REQUIREMENT in repo.requirements
 
 
 def _readroots(
     repo: "localrepo.localrepository",
     phasedefaults: Optional["Phasedefaults"] = None,
 ) -> Tuple[Phaseroots, bool]:
     """Read phase roots from disk
 
     phasedefaults is a list of fn(repo, roots) callable, which are
     executed if the phase roots file does not exist. When phases are
     being initialized on an existing repository, this could be used to
     set selected changesets phase to something else than public.
 
     Return (roots, dirty) where dirty is true if roots differ from
     what is being stored.
     """
     repo = repo.unfiltered()
     dirty = False
     roots = {i: set() for i in allphases}
     to_rev = repo.changelog.index.get_rev
     unknown_msg = b'removing unknown node %s from %i-phase boundary\n'
     try:
         f, pending = txnutil.trypending(repo.root, repo.svfs, b'phaseroots')
         try:
             for line in f:
                 str_phase, hex_node = line.split()
                 phase = int(str_phase)
                 node = bin(hex_node)
                 rev = to_rev(node)
                 if rev is None:
                     repo.ui.debug(unknown_msg % (short(hex_node), phase))
                     dirty = True
                 else:
                     roots[phase].add(rev)
         finally:
             f.close()
     except FileNotFoundError:
         if phasedefaults:
             for f in phasedefaults:
                 roots = f(repo, roots)
         dirty = True
     return roots, dirty
 
 
 def binaryencode(phasemapping: Dict[int, List[bytes]]) -> bytes:
     """encode a 'phase -> nodes' mapping into a binary stream
 
     The revision lists are encoded as (phase, root) pairs.
     """
     binarydata = []
     for phase, nodes in phasemapping.items():
         for head in nodes:
             binarydata.append(_fphasesentry.pack(phase, head))
     return b''.join(binarydata)
 
 
 def binarydecode(stream) -> Dict[int, List[bytes]]:
     """decode a binary stream into a 'phase -> nodes' mapping
 
     The (phase, root) pairs are turned back into a dictionary with
     the phase as index and the aggregated roots of that phase as value."""
     headsbyphase = {i: [] for i in allphases}
     entrysize = _fphasesentry.size
     while True:
         entry = stream.read(entrysize)
         if len(entry) < entrysize:
             if entry:
                 raise error.Abort(_(b'bad phase-heads stream'))
             break
         phase, node = _fphasesentry.unpack(entry)
         headsbyphase[phase].append(node)
     return headsbyphase
 
 
 def _sortedrange_insert(data, idx, rev, t):
     merge_before = False
     if idx:
         r1, t1 = data[idx - 1]
         merge_before = r1[-1] + 1 == rev and t1 == t
     merge_after = False
     if idx < len(data):
         r2, t2 = data[idx]
         merge_after = r2[0] == rev + 1 and t2 == t
 
     if merge_before and merge_after:
         data[idx - 1] = (range(r1[0], r2[-1] + 1), t)
         data.pop(idx)
     elif merge_before:
         data[idx - 1] = (range(r1[0], rev + 1), t)
     elif merge_after:
         data[idx] = (range(rev, r2[-1] + 1), t)
     else:
         data.insert(idx, (range(rev, rev + 1), t))
 
 
 def _sortedrange_split(data, idx, rev, t):
     r1, t1 = data[idx]
     if t == t1:
         return
     t = (t1[0], t[1])
     if len(r1) == 1:
         data.pop(idx)
         _sortedrange_insert(data, idx, rev, t)
     elif r1[0] == rev:
         data[idx] = (range(rev + 1, r1[-1] + 1), t1)
         _sortedrange_insert(data, idx, rev, t)
     elif r1[-1] == rev:
         data[idx] = (range(r1[0], rev), t1)
         _sortedrange_insert(data, idx + 1, rev, t)
     else:
         data[idx : idx + 1] = [
             (range(r1[0], rev), t1),
             (range(rev, rev + 1), t),
             (range(rev + 1, r1[-1] + 1), t1),
         ]
 
 
 def _trackphasechange(data, rev, old, new):
     """add a phase move to the <data> list of ranges
 
     If data is None, nothing happens.
     """
     if data is None:
         return
 
     # If data is empty, create a one-revision range and done
     if not data:
         data.insert(0, (range(rev, rev + 1), (old, new)))
         return
 
     low = 0
     high = len(data)
     t = (old, new)
     while low < high:
         mid = (low + high) // 2
         revs = data[mid][0]
         revs_low = revs[0]
         revs_high = revs[-1]
 
         if rev >= revs_low and rev <= revs_high:
             _sortedrange_split(data, mid, rev, t)
             return
 
         if revs_low == rev + 1:
             if mid and data[mid - 1][0][-1] == rev:
                 _sortedrange_split(data, mid - 1, rev, t)
             else:
                 _sortedrange_insert(data, mid, rev, t)
             return
 
         if revs_high == rev - 1:
             if mid + 1 < len(data) and data[mid + 1][0][0] == rev:
                 _sortedrange_split(data, mid + 1, rev, t)
             else:
                 _sortedrange_insert(data, mid + 1, rev, t)
             return
 
         if revs_low > rev:
             high = mid
         else:
             low = mid + 1
 
     if low == len(data):
         data.append((range(rev, rev + 1), t))
         return
 
     r1, t1 = data[low]
     if r1[0] > rev:
         data.insert(low, (range(rev, rev + 1), t))
     else:
         data.insert(low + 1, (range(rev, rev + 1), t))
 
 
 # consider incrementaly updating the phase set the update set is not bigger
 # than this size
 #
 # Be warned, this number is picked arbitrarily, without any benchmark. It
 # should blindly pickup "small update"
 INCREMENTAL_PHASE_SETS_UPDATE_MAX_UPDATE = 100
 
 
 class phasecache:
     def __init__(
         self,
         repo: "localrepo.localrepository",
         phasedefaults: Optional["Phasedefaults"],
         _load: bool = True,
     ):
         if _load:
             # Cheap trick to allow shallow-copy without copy module
             loaded = _readroots(repo, phasedefaults)
             self._phaseroots: Phaseroots = loaded[0]
             self.dirty: bool = loaded[1]
             self._loadedrevslen = 0
             self._phasesets: PhaseSets = None
 
     def hasnonpublicphases(self, repo: "localrepo.localrepository") -> bool:
         """detect if there are revisions with non-public phase"""
         # XXX deprecate the unused repo argument
         return any(
             revs for phase, revs in self._phaseroots.items() if phase != public
         )
 
     def nonpublicphaseroots(
         self, repo: "localrepo.localrepository"
     ) -> Set[int]:
         """returns the roots of all non-public phases
 
         The roots are not minimized, so if the secret revisions are
         descendants of draft revisions, their roots will still be present.
         """
         repo = repo.unfiltered()
         self._ensure_phase_sets(repo)
         return set().union(
             *[
                 revs
                 for phase, revs in self._phaseroots.items()
                 if phase != public
             ]
         )
 
     def getrevset(
         self,
         repo: "localrepo.localrepository",
         phases: Iterable[int],
         subset: Optional[Any] = None,
     ) -> Any:
         # TODO: finish typing this
         """return a smartset for the given phases"""
         self._ensure_phase_sets(repo.unfiltered())
         phases = set(phases)
         publicphase = public in phases
 
         if publicphase:
             # In this case, phases keeps all the *other* phases.
             phases = set(allphases).difference(phases)
             if not phases:
                 return smartset.fullreposet(repo)
 
         # fast path: _phasesets contains the interesting sets,
         # might only need a union and post-filtering.
         revsneedscopy = False
         if len(phases) == 1:
             [p] = phases
             revs = self._phasesets[p]
             revsneedscopy = True  # Don't modify _phasesets
         else:
             # revs has the revisions in all *other* phases.
             revs = set.union(*[self._phasesets[p] for p in phases])
 
         def _addwdir(wdirsubset, wdirrevs):
             if wdirrev in wdirsubset and repo[None].phase() in phases:
                 if revsneedscopy:
                     wdirrevs = wdirrevs.copy()
                 # The working dir would never be in the # cache, but it was in
                 # the subset being filtered for its phase (or filtered out,
                 # depending on publicphase), so add it to the output to be
                 # included (or filtered out).
                 wdirrevs.add(wdirrev)
             return wdirrevs
 
         if not publicphase:
             if repo.changelog.filteredrevs:
                 revs = revs - repo.changelog.filteredrevs
 
             if subset is None:
                 return smartset.baseset(revs)
             else:
                 revs = _addwdir(subset, revs)
                 return subset & smartset.baseset(revs)
         else:
             if subset is None:
                 subset = smartset.fullreposet(repo)
 
             revs = _addwdir(subset, revs)
 
             if not revs:
                 return subset
             return subset.filter(lambda r: r not in revs)
 
     def copy(self):
         # Shallow copy meant to ensure isolation in
         # advance/retractboundary(), nothing more.
         ph = self.__class__(None, None, _load=False)
         ph._phaseroots = self._phaseroots.copy()
         ph.dirty = self.dirty
         ph._loadedrevslen = self._loadedrevslen
         if self._phasesets is None:
             ph._phasesets = None
         else:
             ph._phasesets = self._phasesets.copy()
         return ph
 
     def replace(self, phcache):
         """replace all values in 'self' with content of phcache"""
         for a in (
             '_phaseroots',
             'dirty',
             '_loadedrevslen',
             '_phasesets',
         ):
             setattr(self, a, getattr(phcache, a))
 
     def _getphaserevsnative(self, repo):
         repo = repo.unfiltered()
         return repo.changelog.computephases(self._phaseroots)
 
     def _computephaserevspure(self, repo):
         repo = repo.unfiltered()
         cl = repo.changelog
         self._phasesets = {phase: set() for phase in allphases}
         lowerroots = set()
         for phase in reversed(trackedphases):
             roots = self._phaseroots[phase]
             if roots:
                 ps = set(cl.descendants(roots))
                 for root in roots:
                     ps.add(root)
                 ps.difference_update(lowerroots)
                 lowerroots.update(ps)
                 self._phasesets[phase] = ps
         self._loadedrevslen = len(cl)
 
     def _ensure_phase_sets(self, repo: "localrepo.localrepository") -> None:
         """ensure phase information is loaded in the object"""
         assert repo.filtername is None
         update = -1
         cl = repo.changelog
         cl_size = len(cl)
         if self._phasesets is None:
             update = 0
         else:
             if cl_size > self._loadedrevslen:
                 # check if an incremental update is worth it.
                 # note we need a tradeoff here because the whole logic is not
                 # stored and implemented in native code nd datastructure.
                 # Otherwise the incremental update woul always be a win.
                 missing = cl_size - self._loadedrevslen
                 if missing <= INCREMENTAL_PHASE_SETS_UPDATE_MAX_UPDATE:
                     update = self._loadedrevslen
                 else:
                     update = 0
 
         if update == 0:
             try:
                 res = self._getphaserevsnative(repo)
                 self._loadedrevslen, self._phasesets = res
             except AttributeError:
                 self._computephaserevspure(repo)
             assert self._loadedrevslen == len(repo.changelog)
         elif update > 0:
             # good candidate for native code
             assert update == self._loadedrevslen
             if self.hasnonpublicphases(repo):
                 start = self._loadedrevslen
                 get_phase = self.phase
                 rev_phases = [0] * missing
                 parents = cl.parentrevs
                 sets = {phase: set() for phase in self._phasesets}
                 for phase, roots in self._phaseroots.items():
                     # XXX should really store the max somewhere
                     for r in roots:
                         if r >= start:
                             rev_phases[r - start] = phase
                 for rev in range(start, cl_size):
                     phase = rev_phases[rev - start]
                     p1, p2 = parents(rev)
                     if p1 == nullrev:
                         p1_phase = public
                     elif p1 >= start:
                         p1_phase = rev_phases[p1 - start]
                     else:
                         p1_phase = max(phase, get_phase(repo, p1))
                     if p2 == nullrev:
                         p2_phase = public
                     elif p2 >= start:
                         p2_phase = rev_phases[p2 - start]
                     else:
                         p2_phase = max(phase, get_phase(repo, p2))
                     phase = max(phase, p1_phase, p2_phase)
                     if phase > public:
                         rev_phases[rev - start] = phase
                         sets[phase].add(rev)
 
                 # Be careful to preserve shallow-copied values: do not update
                 # phaseroots values, replace them.
                 for phase, extra in sets.items():
                     if extra:
                         self._phasesets[phase] = self._phasesets[phase] | extra
             self._loadedrevslen = cl_size
 
     def invalidate(self):
         self._loadedrevslen = 0
         self._phasesets = None
 
     def phase(self, repo: "localrepo.localrepository", rev: int) -> int:
         # We need a repo argument here to be able to build _phasesets
         # if necessary. The repository instance is not stored in
         # phasecache to avoid reference cycles. The changelog instance
         # is not stored because it is a filecache() property and can
         # be replaced without us being notified.
         if rev == nullrev:
             return public
         if rev < nullrev:
             raise ValueError(_(b'cannot lookup negative revision'))
         # double check self._loadedrevslen to avoid an extra method call as
         # python is slow for that.
         if rev >= self._loadedrevslen:
             self._ensure_phase_sets(repo.unfiltered())
         for phase in trackedphases:
             if rev in self._phasesets[phase]:
                 return phase
         return public
 
     def write(self, repo):
         if not self.dirty:
             return
         f = repo.svfs(b'phaseroots', b'w', atomictemp=True, checkambig=True)
         try:
             self._write(repo.unfiltered(), f)
         finally:
             f.close()
 
     def _write(self, repo, fp):
         assert repo.filtername is None
         to_node = repo.changelog.node
         for phase, roots in self._phaseroots.items():
             for r in sorted(roots):
                 h = to_node(r)
                 fp.write(b'%i %s\n' % (phase, hex(h)))
         self.dirty = False
 
     def _updateroots(self, repo, phase, newroots, tr, invalidate=True):
         self._phaseroots[phase] = newroots
         self.dirty = True
         if invalidate:
             self.invalidate()
 
         assert repo.filtername is None
         wrepo = weakref.ref(repo)
 
         def tr_write(fp):
             repo = wrepo()
             assert repo is not None
             self._write(repo, fp)
 
         tr.addfilegenerator(b'phase', (b'phaseroots',), tr_write)
         tr.hookargs[b'phases_moved'] = b'1'
 
     def registernew(self, repo, tr, targetphase, revs):
         repo = repo.unfiltered()
         self._retractboundary(repo, tr, targetphase, [], revs=revs)
         if tr is not None and b'phases' in tr.changes:
             phasetracking = tr.changes[b'phases']
             phase = self.phase
             for rev in sorted(revs):
                 revphase = phase(repo, rev)
                 _trackphasechange(phasetracking, rev, None, revphase)
         repo.invalidatevolatilesets()
 
     def advanceboundary(
         self, repo, tr, targetphase, nodes=None, revs=None, dryrun=None
     ):
         """Set all 'nodes' to phase 'targetphase'
 
         Nodes with a phase lower than 'targetphase' are not affected.
 
         If dryrun is True, no actions will be performed
 
         Returns a set of revs whose phase is changed or should be changed
         """
         if targetphase == public and not self.hasnonpublicphases(repo):
             return set()
         repo = repo.unfiltered()
         cl = repo.changelog
         torev = cl.index.rev
         # Be careful to preserve shallow-copied values: do not update
         # phaseroots values, replace them.
         new_revs = set()
         if revs is not None:
             new_revs.update(revs)
         if nodes is not None:
             new_revs.update(torev(node) for node in nodes)
         if not new_revs:  # bail out early to avoid the loadphaserevs call
             return (
                 set()
             )  # note: why do people call advanceboundary with nothing?
 
         if tr is None:
             phasetracking = None
         else:
             phasetracking = tr.changes.get(b'phases')
 
         affectable_phases = sorted(
             p for p in allphases if p > targetphase and self._phaseroots[p]
         )
         # filter revision already in the right phases
         candidates = new_revs
         new_revs = set()
         self._ensure_phase_sets(repo)
         for phase in affectable_phases:
             found = candidates & self._phasesets[phase]
             new_revs |= found
             candidates -= found
             if not candidates:
                 break
         if not new_revs:
             return set()
 
         # search for affected high phase changesets and roots
         seen = set(new_revs)
         push = heapq.heappush
         pop = heapq.heappop
         parents = cl.parentrevs
         get_phase = self.phase
         changed = {}  # set of revisions to be changed
         # set of root deleted by this path
         delroots = set()
         new_roots = {p: set() for p in affectable_phases}
         new_target_roots = set()
         # revision to walk down
         revs = [-r for r in new_revs]
         heapq.heapify(revs)
         while revs:
             current = -pop(revs)
             current_phase = get_phase(repo, current)
             changed[current] = current_phase
             p1, p2 = parents(current)
             if p1 == nullrev:
                 p1_phase = public
             else:
                 p1_phase = get_phase(repo, p1)
             if p2 == nullrev:
                 p2_phase = public
             else:
                 p2_phase = get_phase(repo, p2)
             # do we have a root ?
             if current_phase != p1_phase and current_phase != p2_phase:
                 # do not record phase, because we could have "duplicated"
                 # roots, were one root is shadowed by the very same roots of an
                 # higher phases
                 delroots.add(current)
             # schedule a walk down if needed
             if p1_phase > targetphase and p1 not in seen:
                 seen.add(p1)
                 push(revs, -p1)
             if p2_phase > targetphase and p2 not in seen:
                 seen.add(p2)
                 push(revs, -p2)
             if p1_phase < targetphase and p2_phase < targetphase:
                 new_target_roots.add(current)
 
         # the last iteration was done with the smallest value
         min_current = current
         # do we have unwalked children that might be new roots
         if (min_current + len(changed)) < len(cl):
             for r in range(min_current, len(cl)):
                 if r in changed:
                     continue
                 phase = get_phase(repo, r)
                 if phase <= targetphase:
                     continue
                 p1, p2 = parents(r)
                 if not (p1 in changed or p2 in changed):
                     continue  # not affected
                 if p1 != nullrev and p1 not in changed:
                     p1_phase = get_phase(repo, p1)
                     if p1_phase == phase:
                         continue  # not a root
                 if p2 != nullrev and p2 not in changed:
                     p2_phase = get_phase(repo, p2)
                     if p2_phase == phase:
                         continue  # not a root
                 new_roots[phase].add(r)
 
         # apply the changes
         if not dryrun:
             for r, p in changed.items():
                 _trackphasechange(phasetracking, r, p, targetphase)
             if targetphase > public:
                 self._phasesets[targetphase].update(changed)
             for phase in affectable_phases:
                 roots = self._phaseroots[phase]
                 removed = roots & delroots
                 if removed or new_roots[phase]:
                     self._phasesets[phase].difference_update(changed)
                     # Be careful to preserve shallow-copied values: do not
                     # update phaseroots values, replace them.
                     final_roots = roots - delroots | new_roots[phase]
                     self._updateroots(
                         repo, phase, final_roots, tr, invalidate=False
                     )
             if new_target_roots:
                 # Thanks for previous filtering, we can't replace existing
                 # roots
                 new_target_roots |= self._phaseroots[targetphase]
                 self._updateroots(
                     repo, targetphase, new_target_roots, tr, invalidate=False
                 )
             repo.invalidatevolatilesets()
         return changed
 
     def retractboundary(self, repo, tr, targetphase, nodes):
         if tr is None:
             phasetracking = None
         else:
             phasetracking = tr.changes.get(b'phases')
         repo = repo.unfiltered()
         retracted = self._retractboundary(repo, tr, targetphase, nodes)
         if retracted and phasetracking is not None:
             for r, old_phase in sorted(retracted.items()):
                 _trackphasechange(phasetracking, r, old_phase, targetphase)
         repo.invalidatevolatilesets()
 
     def _retractboundary(self, repo, tr, targetphase, nodes=None, revs=None):
         if targetphase == public:
             return {}
         if (
             targetphase == internal
             and not supportinternal(repo)
             or targetphase == archived
             and not supportarchived(repo)
         ):
             name = phasenames[targetphase]
             msg = b'this repository does not support the %s phase' % name
             raise error.ProgrammingError(msg)
         assert repo.filtername is None
         cl = repo.changelog
         torev = cl.index.rev
         new_revs = set()
         if revs is not None:
             new_revs.update(revs)
         if nodes is not None:
             new_revs.update(torev(node) for node in nodes)
         if not new_revs:  # bail out early to avoid the loadphaserevs call
             return {}  # note: why do people call retractboundary with nothing ?
 
         if nullrev in new_revs:
             raise error.Abort(_(b'cannot change null revision phase'))
 
         # Filter revision that are already in the right phase
         self._ensure_phase_sets(repo)
         for phase, revs in self._phasesets.items():
             if phase >= targetphase:
                 new_revs -= revs
         if not new_revs:  # all revisions already in the right phases
             return {}
 
         # Compute change in phase roots by walking the graph
         #
         # note: If we had a cheap parent → children mapping we could do
         # something even cheaper/more-bounded
         #
         # The idea would be to walk from item in new_revs stopping at
         # descendant with phases >= target_phase.
         #
         # 1) This detect new_revs that are not new_roots (either already >=
         #    target_phase or reachable though another new_revs
         # 2) This detect replaced current_roots as we reach them
         # 3) This can avoid walking to the tip if we retract over a small
         #    branch.
         #
         # So instead, we do a variation of this, we walk from the smaller new
         # revision to the tip to avoid missing any potential children.
         #
         # The following code would be a good candidate for native code… if only
         # we could knew the phase of a changeset efficiently in native code.
         parents = cl.parentrevs
         phase = self.phase
         new_roots = set()  # roots added by this phases
         changed_revs = {}  # revision affected by this call
         replaced_roots = set()  # older roots replaced by this call
         currentroots = self._phaseroots[targetphase]
         start = min(new_revs)
         end = len(cl)
         rev_phases = [None] * (end - start)
         for r in range(start, end):
 
             # gather information about the current_rev
             r_phase = phase(repo, r)
             p_phase = None  # phase inherited from parents
             p1, p2 = parents(r)
             if p1 >= start:
                 p1_phase = rev_phases[p1 - start]
                 if p1_phase is not None:
                     p_phase = p1_phase
             if p2 >= start:
                 p2_phase = rev_phases[p2 - start]
                 if p2_phase is not None:
                     if p_phase is not None:
                         p_phase = max(p_phase, p2_phase)
                     else:
                         p_phase = p2_phase
 
             # assess the situation
             if r in new_revs and r_phase < targetphase:
                 if p_phase is None or p_phase < targetphase:
                     new_roots.add(r)
                 rev_phases[r - start] = targetphase
                 changed_revs[r] = r_phase
             elif p_phase is None:
                 rev_phases[r - start] = r_phase
             else:
                 if p_phase > r_phase:
                     rev_phases[r - start] = p_phase
                 else:
                     rev_phases[r - start] = r_phase
                 if p_phase == targetphase:
                     if p_phase > r_phase:
                         changed_revs[r] = r_phase
                     elif r in currentroots:
                         replaced_roots.add(r)
             sets = self._phasesets
             sets[targetphase].update(changed_revs)
-            for r, old in changed_revs.items():
-                if old > public:
-                    sets[old].discard(r)
+            if targetphase > draft:
+                for r, old in changed_revs.items():
+                    if old > public:
+                        sets[old].discard(r)
 
         if new_roots:
             assert changed_revs
 
             final_roots = new_roots | currentroots - replaced_roots
             self._updateroots(
                 repo,
                 targetphase,
                 final_roots,
                 tr,
                 invalidate=False,
             )
             if targetphase > 1:
                 retracted = set(changed_revs)
                 for lower_phase in range(1, targetphase):
                     lower_roots = self._phaseroots.get(lower_phase)
                     if lower_roots is None:
                         continue
                     if lower_roots & retracted:
                         simpler_roots = lower_roots - retracted
                         self._updateroots(
                             repo,
                             lower_phase,
                             simpler_roots,
                             tr,
                             invalidate=False,
                         )
             return changed_revs
         else:
             assert not changed_revs
             assert not replaced_roots
             return {}
 
     def register_strip(
         self,
         repo,
         tr,
         strip_rev: int,
     ):
         """announce a strip to the phase cache
 
         Any roots higher than the stripped revision should be dropped.
         """
         for targetphase, roots in list(self._phaseroots.items()):
             filtered = {r for r in roots if r >= strip_rev}
             if filtered:
                 self._updateroots(repo, targetphase, roots - filtered, tr)
         self.invalidate()
 
 
 def advanceboundary(repo, tr, targetphase, nodes, revs=None, dryrun=None):
     """Add nodes to a phase changing other nodes phases if necessary.
 
     This function move boundary *forward* this means that all nodes
     are set in the target phase or kept in a *lower* phase.
 
     Simplify boundary to contains phase roots only.
 
     If dryrun is True, no actions will be performed
 
     Returns a set of revs whose phase is changed or should be changed
     """
     if revs is None:
         revs = []
     phcache = repo._phasecache.copy()
     changes = phcache.advanceboundary(
         repo, tr, targetphase, nodes, revs=revs, dryrun=dryrun
     )
     if not dryrun:
         repo._phasecache.replace(phcache)
     return changes
 
 
 def retractboundary(repo, tr, targetphase, nodes):
     """Set nodes back to a phase changing other nodes phases if
     necessary.
 
     This function move boundary *backward* this means that all nodes
     are set in the target phase or kept in a *higher* phase.
 
     Simplify boundary to contains phase roots only."""
     phcache = repo._phasecache.copy()
     phcache.retractboundary(repo, tr, targetphase, nodes)
     repo._phasecache.replace(phcache)
 
 
 def registernew(repo, tr, targetphase, revs):
     """register a new revision and its phase
 
     Code adding revisions to the repository should use this function to
     set new changeset in their target phase (or higher).
     """
     phcache = repo._phasecache.copy()
     phcache.registernew(repo, tr, targetphase, revs)
     repo._phasecache.replace(phcache)
 
 
 def listphases(repo: "localrepo.localrepository") -> Dict[bytes, bytes]:
     """List phases root for serialization over pushkey"""
     # Use ordered dictionary so behavior is deterministic.
     keys = util.sortdict()
     value = b'%i' % draft
     cl = repo.unfiltered().changelog
     to_node = cl.node
     for root in repo._phasecache._phaseroots[draft]:
         if repo._phasecache.phase(repo, root) <= draft:
             keys[hex(to_node(root))] = value
 
     if repo.publishing():
         # Add an extra data to let remote know we are a publishing
         # repo. Publishing repo can't just pretend they are old repo.
         # When pushing to a publishing repo, the client still need to
         # push phase boundary
         #
         # Push do not only push changeset. It also push phase data.
         # New phase data may apply to common changeset which won't be
         # push (as they are common). Here is a very simple example:
         #
         # 1) repo A push changeset X as draft to repo B
         # 2) repo B make changeset X public
         # 3) repo B push to repo A. X is not pushed but the data that
         #    X as now public should
         #
         # The server can't handle it on it's own as it has no idea of
         # client phase data.
         keys[b'publishing'] = b'True'
     return keys
 
 
 def pushphase(
     repo: "localrepo.localrepository",
     nhex: bytes,
     oldphasestr: bytes,
     newphasestr: bytes,
 ) -> bool:
     """List phases root for serialization over pushkey"""
     repo = repo.unfiltered()
     with repo.lock():
         currentphase = repo[nhex].phase()
         newphase = abs(int(newphasestr))  # let's avoid negative index surprise
         oldphase = abs(int(oldphasestr))  # let's avoid negative index surprise
         if currentphase == oldphase and newphase < oldphase:
             with repo.transaction(b'pushkey-phase') as tr:
                 advanceboundary(repo, tr, newphase, [bin(nhex)])
             return True
         elif currentphase == newphase:
             # raced, but got correct result
             return True
         else:
             return False
 
 
 def subsetphaseheads(repo, subset):
     """Finds the phase heads for a subset of a history
 
     Returns a list indexed by phase number where each item is a list of phase
     head nodes.
     """
     cl = repo.changelog
 
     headsbyphase = {i: [] for i in allphases}
     for phase in allphases:
         revset = b"heads(%%ln & _phase(%d))" % phase
         headsbyphase[phase] = [cl.node(r) for r in repo.revs(revset, subset)]
     return headsbyphase
 
 
 def updatephases(repo, trgetter, headsbyphase):
     """Updates the repo with the given phase heads"""
     # Now advance phase boundaries of all phases
     #
     # run the update (and fetch transaction) only if there are actually things
     # to update. This avoid creating empty transaction during no-op operation.
 
     for phase in allphases:
         revset = b'%ln - _phase(%s)'
         heads = [c.node() for c in repo.set(revset, headsbyphase[phase], phase)]
         if heads:
             advanceboundary(repo, trgetter(), phase, heads)
 
 
 def analyzeremotephases(repo, subset, roots):
     """Compute phases heads and root in a subset of node from root dict
 
     * subset is heads of the subset
     * roots is {<nodeid> => phase} mapping. key and value are string.
 
     Accept unknown element input
     """
     repo = repo.unfiltered()
     # build list from dictionary
     draftroots = []
     has_node = repo.changelog.index.has_node  # to filter unknown nodes
     for nhex, phase in roots.items():
         if nhex == b'publishing':  # ignore data related to publish option
             continue
         node = bin(nhex)
         phase = int(phase)
         if phase == public:
             if node != repo.nullid:
                 repo.ui.warn(
                     _(
                         b'ignoring inconsistent public root'
                         b' from remote: %s\n'
                     )
                     % nhex
                 )
         elif phase == draft:
             if has_node(node):
                 draftroots.append(node)
         else:
             repo.ui.warn(
                 _(b'ignoring unexpected root from remote: %i %s\n')
                 % (phase, nhex)
             )
     # compute heads
     publicheads = newheads(repo, subset, draftroots)
     return publicheads, draftroots
 
 
 class remotephasessummary:
     """summarize phase information on the remote side
 
     :publishing: True is the remote is publishing
     :publicheads: list of remote public phase heads (nodes)
     :draftheads: list of remote draft phase heads (nodes)
     :draftroots: list of remote draft phase root (nodes)
     """
 
     def __init__(self, repo, remotesubset, remoteroots):
         unfi = repo.unfiltered()
         self._allremoteroots = remoteroots
 
         self.publishing = remoteroots.get(b'publishing', False)
 
         ana = analyzeremotephases(repo, remotesubset, remoteroots)
         self.publicheads, self.draftroots = ana
         # Get the list of all "heads" revs draft on remote
         dheads = unfi.set(b'heads(%ln::%ln)', self.draftroots, remotesubset)
         self.draftheads = [c.node() for c in dheads]
 
 
 def newheads(repo, heads, roots):
     """compute new head of a subset minus another
 
     * `heads`: define the first subset
     * `roots`: define the second we subtract from the first"""
     # prevent an import cycle
     # phases > dagop > patch > copies > scmutil > obsolete > obsutil > phases
     from . import dagop
 
     repo = repo.unfiltered()
     cl = repo.changelog
     rev = cl.index.get_rev
     if not roots:
         return heads
     if not heads or heads == [repo.nullid]:
         return []
     # The logic operated on revisions, convert arguments early for convenience
     new_heads = {rev(n) for n in heads if n != repo.nullid}
     roots = [rev(n) for n in roots]
     # compute the area we need to remove
     affected_zone = repo.revs(b"(%ld::%ld)", roots, new_heads)
     # heads in the area are no longer heads
     new_heads.difference_update(affected_zone)
     # revisions in the area have children outside of it,
     # They might be new heads
     candidates = repo.revs(
         b"parents(%ld + (%ld and merge())) and not null", roots, affected_zone
     )
     candidates -= affected_zone
     if new_heads or candidates:
         # remove candidate that are ancestors of other heads
         new_heads.update(candidates)
         prunestart = repo.revs(b"parents(%ld) and not null", new_heads)
         pruned = dagop.reachableroots(repo, candidates, prunestart)
         new_heads.difference_update(pruned)
 
     return pycompat.maplist(cl.node, sorted(new_heads))
 
 
 def newcommitphase(ui: "uimod.ui") -> int:
     """helper to get the target phase of new commit
 
     Handle all possible values for the phases.new-commit options.
 
     """
     v = ui.config(b'phases', b'new-commit')
     try:
         return phasenumber2[v]
     except KeyError:
         raise error.ConfigError(
             _(b"phases.new-commit: not a valid phase name ('%s')") % v
         )
 
 
 def hassecret(repo: "localrepo.localrepository") -> bool:
     """utility function that check if a repo have any secret changeset."""
     return bool(repo._phasecache._phaseroots[secret])
 
 
 def preparehookargs(
     node: bytes,
     old: Optional[int],
     new: Optional[int],
 ) -> Dict[bytes, bytes]:
     if old is None:
         old = b''
     else:
         old = phasenames[old]
     return {b'node': node, b'oldphase': old, b'phase': phasenames[new]}