absorb: import extension from Facebook's hg-experimental

absorb is a wicked-fast command to use blame information to
automatically amend edits to the correct draft revision. Originally
written by Jun Wu, this import is hgext3rd/absorb/__init__.py with:
 * the `testedwith` value changed
 * the linelog import updated
 * some missing configitems registered
 * some imports reordered per check-code.py
 * some missing __future__ imports added per check-code.py

Differential Revision: https://phab.mercurial-scm.org/D3991

hgext/absorb.py

@@ -0,0 +1,1041 b''
+# absorb.py
+#
+# Copyright 2016 Facebook, Inc.
+#
+# This software may be used and distributed according to the terms of the
+# GNU General Public License version 2 or any later version.
+
+"""apply working directory changes to changesets (EXPERIMENTAL)
+
+The absorb extension provides a command to use annotate information to
+amend modified chunks into the corresponding non-public changesets.
+
+::
+
+    [absorb]
+    # only check 50 recent non-public changesets at most
+    maxstacksize = 50
+    # whether to add noise to new commits to avoid obsolescence cycle
+    addnoise = 1
+    # make `amend --correlated` a shortcut to the main command
+    amendflag = correlated
+
+    [color]
+    absorb.node = blue bold
+    absorb.path = bold
+"""
+
+from __future__ import absolute_import
+
+import collections
+
+from mercurial.i18n import _
+from mercurial import (
+    cmdutil,
+    commands,
+    context,
+    crecord,
+    error,
+    extensions,
+    linelog,
+    mdiff,
+    node,
+    obsolete,
+    patch,
+    phases,
+    registrar,
+    repair,
+    scmutil,
+    util,
+)
+from mercurial.utils import (
+    stringutil,
+)
+
+# Note for extension authors: ONLY specify testedwith = 'ships-with-hg-core' for
+# extensions which SHIP WITH MERCURIAL. Non-mainline extensions should
+# be specifying the version(s) of Mercurial they are tested with, or
+# leave the attribute unspecified.
+testedwith = 'ships-with-hg-core'
+
+cmdtable = {}
+command = registrar.command(cmdtable)
+
+configtable = {}
+configitem = registrar.configitem(configtable)
+
+configitem('absorb', 'addnoise', default=True)
+configitem('absorb', 'amendflag', default=None)
+configitem('absorb', 'maxstacksize', default=50)
+
+colortable = {
+    'absorb.node': 'blue bold',
+    'absorb.path': 'bold',
+}
+
+defaultdict = collections.defaultdict
+
+class nullui(object):
+    """blank ui object doing nothing"""
+    debugflag = False
+    verbose = False
+    quiet = True
+
+    def __getitem__(name):
+        def nullfunc(*args, **kwds):
+            return
+        return nullfunc
+
+class emptyfilecontext(object):
+    """minimal filecontext representing an empty file"""
+    def data(self):
+        return ''
+
+    def node(self):
+        return node.nullid
+
+def uniq(lst):
+    """list -> list. remove duplicated items without changing the order"""
+    seen = set()
+    result = []
+    for x in lst:
+        if x not in seen:
+            seen.add(x)
+            result.append(x)
+    return result
+
+def getdraftstack(headctx, limit=None):
+    """(ctx, int?) -> [ctx]. get a linear stack of non-public changesets.
+
+    changesets are sorted in topo order, oldest first.
+    return at most limit items, if limit is a positive number.
+
+    merges are considered as non-draft as well. i.e. every commit
+    returned has and only has 1 parent.
+    """
+    ctx = headctx
+    result = []
+    while ctx.phase() != phases.public:
+        if limit and len(result) >= limit:
+            break
+        parents = ctx.parents()
+        if len(parents) != 1:
+            break
+        result.append(ctx)
+        ctx = parents[0]
+    result.reverse()
+    return result
+
+def getfilestack(stack, path, seenfctxs=set()):
+    """([ctx], str, set) -> [fctx], {ctx: fctx}
+
+    stack is a list of contexts, from old to new. usually they are what
+    "getdraftstack" returns.
+
+    follows renames, but not copies.
+
+    seenfctxs is a set of filecontexts that will be considered "immutable".
+    they are usually what this function returned in earlier calls, useful
+    to avoid issues that a file was "moved" to multiple places and was then
+    modified differently, like: "a" was copied to "b", "a" was also copied to
+    "c" and then "a" was deleted, then both "b" and "c" were "moved" from "a"
+    and we enforce only one of them to be able to affect "a"'s content.
+
+    return an empty list and an empty dict, if the specified path does not
+    exist in stack[-1] (the top of the stack).
+
+    otherwise, return a list of de-duplicated filecontexts, and the map to
+    convert ctx in the stack to fctx, for possible mutable fctxs. the first item
+    of the list would be outside the stack and should be considered immutable.
+    the remaining items are within the stack.
+
+    for example, given the following changelog and corresponding filelog
+    revisions:
+
+      changelog: 3----4----5----6----7
+      filelog:   x    0----1----1----2 (x: no such file yet)
+
+    - if stack = [5, 6, 7], returns ([0, 1, 2], {5: 1, 6: 1, 7: 2})
+    - if stack = [3, 4, 5], returns ([e, 0, 1], {4: 0, 5: 1}), where "e" is a
+      dummy empty filecontext.
+    - if stack = [2], returns ([], {})
+    - if stack = [7], returns ([1, 2], {7: 2})
+    - if stack = [6, 7], returns ([1, 2], {6: 1, 7: 2}), although {6: 1} can be
+      removed, since 1 is immutable.
+    """
+    assert stack
+
+    if path not in stack[-1]:
+        return [], {}
+
+    fctxs = []
+    fctxmap = {}
+
+    pctx = stack[0].p1() # the public (immutable) ctx we stop at
+    for ctx in reversed(stack):
+        if path not in ctx: # the file is added in the next commit
+            pctx = ctx
+            break
+        fctx = ctx[path]
+        fctxs.append(fctx)
+        if fctx in seenfctxs: # treat fctx as the immutable one
+            pctx = None # do not add another immutable fctx
+            break
+        fctxmap[ctx] = fctx # only for mutable fctxs
+        renamed = fctx.renamed()
+        if renamed:
+            path = renamed[0] # follow rename
+            if path in ctx: # but do not follow copy
+                pctx = ctx.p1()
+                break
+
+    if pctx is not None: # need an extra immutable fctx
+        if path in pctx:
+            fctxs.append(pctx[path])
+        else:
+            fctxs.append(emptyfilecontext())
+
+    fctxs.reverse()
+    # note: we rely on a property of hg: filerev is not reused for linear
+    # history. i.e. it's impossible to have:
+    #   changelog:  4----5----6 (linear, no merges)
+    #   filelog:    1----2----1
+    #                         ^ reuse filerev (impossible)
+    # because parents are part of the hash. if that's not true, we need to
+    # remove uniq and find a different way to identify fctxs.
+    return uniq(fctxs), fctxmap
+
+class overlaystore(patch.filestore):
+    """read-only, hybrid store based on a dict and ctx.
+    memworkingcopy: {path: content}, overrides file contents.
+    """
+    def __init__(self, basectx, memworkingcopy):
+        self.basectx = basectx
+        self.memworkingcopy = memworkingcopy
+
+    def getfile(self, path):
+        """comply with mercurial.patch.filestore.getfile"""
+        if path not in self.basectx:
+            return None, None, None
+        fctx = self.basectx[path]
+        if path in self.memworkingcopy:
+            content = self.memworkingcopy[path]
+        else:
+            content = fctx.data()
+        mode = (fctx.islink(), fctx.isexec())
+        renamed = fctx.renamed() # False or (path, node)
+        return content, mode, (renamed and renamed[0])
+
+def overlaycontext(memworkingcopy, ctx, parents=None, extra=None):
+    """({path: content}, ctx, (p1node, p2node)?, {}?) -> memctx
+    memworkingcopy overrides file contents.
+    """
+    # parents must contain 2 items: (node1, node2)
+    if parents is None:
+        parents = ctx.repo().changelog.parents(ctx.node())
+    if extra is None:
+        extra = ctx.extra()
+    date = ctx.date()
+    desc = ctx.description()
+    user = ctx.user()
+    files = set(ctx.files()).union(memworkingcopy.iterkeys())
+    store = overlaystore(ctx, memworkingcopy)
+    return context.memctx(
+        repo=ctx.repo(), parents=parents, text=desc,
+        files=files, filectxfn=store, user=user, date=date,
+        branch=None, extra=extra)
+
+class filefixupstate(object):
+    """state needed to apply fixups to a single file
+
+    internally, it keeps file contents of several revisions and a linelog.
+
+    the linelog uses odd revision numbers for original contents (fctxs passed
+    to __init__), and even revision numbers for fixups, like:
+
+        linelog rev 1: self.fctxs[0] (from an immutable "public" changeset)
+        linelog rev 2: fixups made to self.fctxs[0]
+        linelog rev 3: self.fctxs[1] (a child of fctxs[0])
+        linelog rev 4: fixups made to self.fctxs[1]
+        ...
+
+    a typical use is like:
+
+        1. call diffwith, to calculate self.fixups
+        2. (optionally), present self.fixups to the user, or change it
+        3. call apply, to apply changes
+        4. read results from "finalcontents", or call getfinalcontent
+    """
+
+    def __init__(self, fctxs, ui=None, opts=None):
+        """([fctx], ui or None) -> None
+
+        fctxs should be linear, and sorted by topo order - oldest first.
+        fctxs[0] will be considered as "immutable" and will not be changed.
+        """
+        self.fctxs = fctxs
+        self.ui = ui or nullui()
+        self.opts = opts or {}
+
+        # following fields are built from fctxs. they exist for perf reason
+        self.contents = [f.data() for f in fctxs]
+        self.contentlines = map(mdiff.splitnewlines, self.contents)
+        self.linelog = self._buildlinelog()
+        if self.ui.debugflag:
+            assert self._checkoutlinelog() == self.contents
+
+        # following fields will be filled later
+        self.chunkstats = [0, 0] # [adopted, total : int]
+        self.targetlines = [] # [str]
+        self.fixups = [] # [(linelog rev, a1, a2, b1, b2)]
+        self.finalcontents = [] # [str]
+
+    def diffwith(self, targetfctx, showchanges=False):
+        """calculate fixups needed by examining the differences between
+        self.fctxs[-1] and targetfctx, chunk by chunk.
+
+        targetfctx is the target state we move towards. we may or may not be
+        able to get there because not all modified chunks can be amended into
+        a non-public fctx unambiguously.
+
+        call this only once, before apply().
+
+        update self.fixups, self.chunkstats, and self.targetlines.
+        """
+        a = self.contents[-1]
+        alines = self.contentlines[-1]
+        b = targetfctx.data()
+        blines = mdiff.splitnewlines(b)
+        self.targetlines = blines
+
+        self.linelog.annotate(self.linelog.maxrev)
+        annotated = self.linelog.annotateresult # [(linelog rev, linenum)]
+        assert len(annotated) == len(alines)
+        # add a dummy end line to make insertion at the end easier
+        if annotated:
+            dummyendline = (annotated[-1][0], annotated[-1][1] + 1)
+            annotated.append(dummyendline)
+
+        # analyse diff blocks
+        for chunk in self._alldiffchunks(a, b, alines, blines):
+            newfixups = self._analysediffchunk(chunk, annotated)
+            self.chunkstats[0] += bool(newfixups) # 1 or 0
+            self.chunkstats[1] += 1
+            self.fixups += newfixups
+            if showchanges:
+                self._showchanges(alines, blines, chunk, newfixups)
+
+    def apply(self):
+        """apply self.fixups. update self.linelog, self.finalcontents.
+
+        call this only once, before getfinalcontent(), after diffwith().
+        """
+        # the following is unnecessary, as it's done by "diffwith":
+        #   self.linelog.annotate(self.linelog.maxrev)
+        for rev, a1, a2, b1, b2 in reversed(self.fixups):
+            blines = self.targetlines[b1:b2]
+            if self.ui.debugflag:
+                idx = (max(rev - 1, 0)) // 2
+                self.ui.write(_('%s: chunk %d:%d -> %d lines\n')
+                              % (node.short(self.fctxs[idx].node()),
+                                 a1, a2, len(blines)))
+            self.linelog.replacelines(rev, a1, a2, b1, b2)
+        if self.opts.get('edit_lines', False):
+            self.finalcontents = self._checkoutlinelogwithedits()
+        else:
+            self.finalcontents = self._checkoutlinelog()
+
+    def getfinalcontent(self, fctx):
+        """(fctx) -> str. get modified file content for a given filecontext"""
+        idx = self.fctxs.index(fctx)
+        return self.finalcontents[idx]
+
+    def _analysediffchunk(self, chunk, annotated):
+        """analyse a different chunk and return new fixups found
+
+        return [] if no lines from the chunk can be safely applied.
+
+        the chunk (or lines) cannot be safely applied, if, for example:
+          - the modified (deleted) lines belong to a public changeset
+            (self.fctxs[0])
+          - the chunk is a pure insertion and the adjacent lines (at most 2
+            lines) belong to different non-public changesets, or do not belong
+            to any non-public changesets.
+          - the chunk is modifying lines from different changesets.
+            in this case, if the number of lines deleted equals to the number
+            of lines added, assume it's a simple 1:1 map (could be wrong).
+            otherwise, give up.
+          - the chunk is modifying lines from a single non-public changeset,
+            but other revisions touch the area as well. i.e. the lines are
+            not continuous as seen from the linelog.
+        """
+        a1, a2, b1, b2 = chunk
+        # find involved indexes from annotate result
+        involved = annotated[a1:a2]
+        if not involved and annotated: # a1 == a2 and a is not empty
+            # pure insertion, check nearby lines. ignore lines belong
+            # to the public (first) changeset (i.e. annotated[i][0] == 1)
+            nearbylinenums = set([a2, max(0, a1 - 1)])
+            involved = [annotated[i]
+                        for i in nearbylinenums if annotated[i][0] != 1]
+        involvedrevs = list(set(r for r, l in involved))
+        newfixups = []
+        if len(involvedrevs) == 1 and self._iscontinuous(a1, a2 - 1, True):
+            # chunk belongs to a single revision
+            rev = involvedrevs[0]
+            if rev > 1:
+                fixuprev = rev + 1
+                newfixups.append((fixuprev, a1, a2, b1, b2))
+        elif a2 - a1 == b2 - b1 or b1 == b2:
+            # 1:1 line mapping, or chunk was deleted
+            for i in xrange(a1, a2):
+                rev, linenum = annotated[i]
+                if rev > 1:
+                    if b1 == b2: # deletion, simply remove that single line
+                        nb1 = nb2 = 0
+                    else: # 1:1 line mapping, change the corresponding rev
+                        nb1 = b1 + i - a1
+                        nb2 = nb1 + 1
+                    fixuprev = rev + 1
+                    newfixups.append((fixuprev, i, i + 1, nb1, nb2))
+        return self._optimizefixups(newfixups)
+
+    @staticmethod
+    def _alldiffchunks(a, b, alines, blines):
+        """like mdiff.allblocks, but only care about differences"""
+        blocks = mdiff.allblocks(a, b, lines1=alines, lines2=blines)
+        for chunk, btype in blocks:
+            if btype != '!':
+                continue
+            yield chunk
+
+    def _buildlinelog(self):
+        """calculate the initial linelog based on self.content{,line}s.
+        this is similar to running a partial "annotate".
+        """
+        llog = linelog.linelog()
+        a, alines = '', []
+        for i in xrange(len(self.contents)):
+            b, blines = self.contents[i], self.contentlines[i]
+            llrev = i * 2 + 1
+            chunks = self._alldiffchunks(a, b, alines, blines)
+            for a1, a2, b1, b2 in reversed(list(chunks)):
+                llog.replacelines(llrev, a1, a2, b1, b2)
+            a, alines = b, blines
+        return llog
+
+    def _checkoutlinelog(self):
+        """() -> [str]. check out file contents from linelog"""
+        contents = []
+        for i in xrange(len(self.contents)):
+            rev = (i + 1) * 2
+            self.linelog.annotate(rev)
+            content = ''.join(map(self._getline, self.linelog.annotateresult))
+            contents.append(content)
+        return contents
+
+    def _checkoutlinelogwithedits(self):
+        """() -> [str]. prompt all lines for edit"""
+        alllines = self.linelog.getalllines()
+        # header
+        editortext = (_('HG: editing %s\nHG: "y" means the line to the right '
+                        'exists in the changeset to the top\nHG:\n')
+                      % self.fctxs[-1].path())
+        # [(idx, fctx)]. hide the dummy emptyfilecontext
+        visiblefctxs = [(i, f)
+                        for i, f in enumerate(self.fctxs)
+                        if not isinstance(f, emptyfilecontext)]
+        for i, (j, f) in enumerate(visiblefctxs):
+            editortext += (_('HG: %s/%s %s %s\n') %
+                           ('|' * i, '-' * (len(visiblefctxs) - i + 1),
+                            node.short(f.node()),
+                            f.description().split('\n',1)[0]))
+        editortext += _('HG: %s\n') % ('|' * len(visiblefctxs))
+        # figure out the lifetime of a line, this is relatively inefficient,
+        # but probably fine
+        lineset = defaultdict(lambda: set()) # {(llrev, linenum): {llrev}}
+        for i, f in visiblefctxs:
+            self.linelog.annotate((i + 1) * 2)
+            for l in self.linelog.annotateresult:
+                lineset[l].add(i)
+        # append lines
+        for l in alllines:
+            editortext += ('    %s : %s' %
+                           (''.join([('y' if i in lineset[l] else ' ')
+                                     for i, _f in visiblefctxs]),
+                            self._getline(l)))
+        # run editor
+        editedtext = self.ui.edit(editortext, '', action='absorb')
+        if not editedtext:
+            raise error.Abort(_('empty editor text'))
+        # parse edited result
+        contents = ['' for i in self.fctxs]
+        leftpadpos = 4
+        colonpos = leftpadpos + len(visiblefctxs) + 1
+        for l in mdiff.splitnewlines(editedtext):
+            if l.startswith('HG:'):
+                continue
+            if l[colonpos - 1:colonpos + 2] != ' : ':
+                raise error.Abort(_('malformed line: %s') % l)
+            linecontent = l[colonpos + 2:]
+            for i, ch in enumerate(l[leftpadpos:colonpos - 1]):
+                if ch == 'y':
+                    contents[visiblefctxs[i][0]] += linecontent
+        # chunkstats is hard to calculate if anything changes, therefore
+        # set them to just a simple value (1, 1).
+        if editedtext != editortext:
+            self.chunkstats = [1, 1]
+        return contents
+
+    def _getline(self, lineinfo):
+        """((rev, linenum)) -> str. convert rev+line number to line content"""
+        rev, linenum = lineinfo
+        if rev & 1: # odd: original line taken from fctxs
+            return self.contentlines[rev // 2][linenum]
+        else: # even: fixup line from targetfctx
+            return self.targetlines[linenum]
+
+    def _iscontinuous(self, a1, a2, closedinterval=False):
+        """(a1, a2 : int) -> bool
+
+        check if these lines are continuous. i.e. no other insertions or
+        deletions (from other revisions) among these lines.
+
+        closedinterval decides whether a2 should be included or not. i.e. is
+        it [a1, a2), or [a1, a2] ?
+        """
+        if a1 >= a2:
+            return True
+        llog = self.linelog
+        offset1 = llog.getoffset(a1)
+        offset2 = llog.getoffset(a2) + int(closedinterval)
+        linesinbetween = llog.getalllines(offset1, offset2)
+        return len(linesinbetween) == a2 - a1 + int(closedinterval)
+
+    def _optimizefixups(self, fixups):
+        """[(rev, a1, a2, b1, b2)] -> [(rev, a1, a2, b1, b2)].
+        merge adjacent fixups to make them less fragmented.
+        """
+        result = []
+        pcurrentchunk = [[-1, -1, -1, -1, -1]]
+
+        def pushchunk():
+            if pcurrentchunk[0][0] != -1:
+                result.append(tuple(pcurrentchunk[0]))
+
+        for i, chunk in enumerate(fixups):
+            rev, a1, a2, b1, b2 = chunk
+            lastrev = pcurrentchunk[0][0]
+            lasta2 = pcurrentchunk[0][2]
+            lastb2 = pcurrentchunk[0][4]
+            if (a1 == lasta2 and b1 == lastb2 and rev == lastrev and
+                    self._iscontinuous(max(a1 - 1, 0), a1)):
+                # merge into currentchunk
+                pcurrentchunk[0][2] = a2
+                pcurrentchunk[0][4] = b2
+            else:
+                pushchunk()
+                pcurrentchunk[0] = list(chunk)
+        pushchunk()
+        return result
+
+    def _showchanges(self, alines, blines, chunk, fixups):
+        ui = self.ui
+
+        def label(line, label):
+            if line.endswith('\n'):
+                line = line[:-1]
+            return ui.label(line, label)
+
+        # this is not optimized for perf but _showchanges only gets executed
+        # with an extra command-line flag.
+        a1, a2, b1, b2 = chunk
+        aidxs, bidxs = [0] * (a2 - a1), [0] * (b2 - b1)
+        for idx, fa1, fa2, fb1, fb2 in fixups:
+            for i in xrange(fa1, fa2):
+                aidxs[i - a1] = (max(idx, 1) - 1) // 2
+            for i in xrange(fb1, fb2):
+                bidxs[i - b1] = (max(idx, 1) - 1) // 2
+
+        buf = [] # [(idx, content)]
+        buf.append((0, label('@@ -%d,%d +%d,%d @@'
+                             % (a1, a2 - a1, b1, b2 - b1), 'diff.hunk')))
+        buf += [(aidxs[i - a1], label('-' + alines[i], 'diff.deleted'))
+                for i in xrange(a1, a2)]
+        buf += [(bidxs[i - b1], label('+' + blines[i], 'diff.inserted'))
+                for i in xrange(b1, b2)]
+        for idx, line in buf:
+            shortnode = idx and node.short(self.fctxs[idx].node()) or ''
+            ui.write(ui.label(shortnode[0:7].ljust(8), 'absorb.node') +
+                     line + '\n')
+
+class fixupstate(object):
+    """state needed to run absorb
+
+    internally, it keeps paths and filefixupstates.
+
+    a typical use is like filefixupstates:
+
+        1. call diffwith, to calculate fixups
+        2. (optionally), present fixups to the user, or edit fixups
+        3. call apply, to apply changes to memory
+        4. call commit, to commit changes to hg database
+    """
+
+    def __init__(self, stack, ui=None, opts=None):
+        """([ctx], ui or None) -> None
+
+        stack: should be linear, and sorted by topo order - oldest first.
+        all commits in stack are considered mutable.
+        """
+        assert stack
+        self.ui = ui or nullui()
+        self.opts = opts or {}
+        self.stack = stack
+        self.repo = stack[-1].repo().unfiltered()
+
+        # following fields will be filled later
+        self.paths = [] # [str]
+        self.status = None # ctx.status output
+        self.fctxmap = {} # {path: {ctx: fctx}}
+        self.fixupmap = {} # {path: filefixupstate}
+        self.replacemap = {} # {oldnode: newnode or None}
+        self.finalnode = None # head after all fixups
+
+    def diffwith(self, targetctx, match=None, showchanges=False):
+        """diff and prepare fixups. update self.fixupmap, self.paths"""
+        # only care about modified files
+        self.status = self.stack[-1].status(targetctx, match)
+        self.paths = []
+        # but if --edit-lines is used, the user may want to edit files
+        # even if they are not modified
+        editopt = self.opts.get('edit_lines')
+        if not self.status.modified and editopt and match:
+            interestingpaths = match.files()
+        else:
+            interestingpaths = self.status.modified
+        # prepare the filefixupstate
+        seenfctxs = set()
+        # sorting is necessary to eliminate ambiguity for the "double move"
+        # case: "hg cp A B; hg cp A C; hg rm A", then only "B" can affect "A".
+        for path in sorted(interestingpaths):
+            if self.ui.debugflag:
+                self.ui.write(_('calculating fixups for %s\n') % path)
+            targetfctx = targetctx[path]
+            fctxs, ctx2fctx = getfilestack(self.stack, path, seenfctxs)
+            # ignore symbolic links or binary, or unchanged files
+            if any(f.islink() or stringutil.binary(f.data())
+                   for f in [targetfctx] + fctxs
+                   if not isinstance(f, emptyfilecontext)):
+                continue
+            if targetfctx.data() == fctxs[-1].data() and not editopt:
+                continue
+            seenfctxs.update(fctxs[1:])
+            self.fctxmap[path] = ctx2fctx
+            fstate = filefixupstate(fctxs, ui=self.ui, opts=self.opts)
+            if showchanges:
+                colorpath = self.ui.label(path, 'absorb.path')
+                header = 'showing changes for ' + colorpath
+                self.ui.write(header + '\n')
+            fstate.diffwith(targetfctx, showchanges=showchanges)
+            self.fixupmap[path] = fstate
+            self.paths.append(path)
+
+    def apply(self):
+        """apply fixups to individual filefixupstates"""
+        for path, state in self.fixupmap.iteritems():
+            if self.ui.debugflag:
+                self.ui.write(_('applying fixups to %s\n') % path)
+            state.apply()
+
+    @property
+    def chunkstats(self):
+        """-> {path: chunkstats}. collect chunkstats from filefixupstates"""
+        return dict((path, state.chunkstats)
+                    for path, state in self.fixupmap.iteritems())
+
+    def commit(self):
+        """commit changes. update self.finalnode, self.replacemap"""
+        with self.repo.wlock(), self.repo.lock():
+            with self.repo.transaction('absorb') as tr:
+                self._commitstack()
+                self._movebookmarks(tr)
+                if self.repo['.'].node() in self.replacemap:
+                    self._moveworkingdirectoryparent()
+                if self._useobsolete:
+                    self._obsoleteoldcommits()
+            if not self._useobsolete: # strip must be outside transactions
+                self._stripoldcommits()
+        return self.finalnode
+
+    def printchunkstats(self):
+        """print things like '1 of 2 chunk(s) applied'"""
+        ui = self.ui
+        chunkstats = self.chunkstats
+        if ui.verbose:
+            # chunkstats for each file
+            for path, stat in chunkstats.iteritems():
+                if stat[0]:
+                    ui.write(_('%s: %d of %d chunk(s) applied\n')
+                             % (path, stat[0], stat[1]))
+        elif not ui.quiet:
+            # a summary for all files
+            stats = chunkstats.values()
+            applied, total = (sum(s[i] for s in stats) for i in (0, 1))
+            ui.write(_('%d of %d chunk(s) applied\n') % (applied, total))
+
+    def _commitstack(self):
+        """make new commits. update self.finalnode, self.replacemap.
+        it is splitted from "commit" to avoid too much indentation.
+        """
+        # last node (20-char) committed by us
+        lastcommitted = None
+        # p1 which overrides the parent of the next commit, "None" means use
+        # the original parent unchanged
+        nextp1 = None
+        for ctx in self.stack:
+            memworkingcopy = self._getnewfilecontents(ctx)
+            if not memworkingcopy and not lastcommitted:
+                # nothing changed, nothing commited
+                nextp1 = ctx
+                continue
+            msg = ''
+            if self._willbecomenoop(memworkingcopy, ctx, nextp1):
+                # changeset is no longer necessary
+                self.replacemap[ctx.node()] = None
+                msg = _('became empty and was dropped')
+            else:
+                # changeset needs re-commit
+                nodestr = self._commitsingle(memworkingcopy, ctx, p1=nextp1)
+                lastcommitted = self.repo[nodestr]
+                nextp1 = lastcommitted
+                self.replacemap[ctx.node()] = lastcommitted.node()
+                if memworkingcopy:
+                    msg = _('%d file(s) changed, became %s') % (
+                        len(memworkingcopy), self._ctx2str(lastcommitted))
+                else:
+                    msg = _('became %s') % self._ctx2str(lastcommitted)
+            if self.ui.verbose and msg:
+                self.ui.write(_('%s: %s\n') % (self._ctx2str(ctx), msg))
+        self.finalnode = lastcommitted and lastcommitted.node()
+
+    def _ctx2str(self, ctx):
+        if self.ui.debugflag:
+            return ctx.hex()
+        else:
+            return node.short(ctx.node())
+
+    def _getnewfilecontents(self, ctx):
+        """(ctx) -> {path: str}
+
+        fetch file contents from filefixupstates.
+        return the working copy overrides - files different from ctx.
+        """
+        result = {}
+        for path in self.paths:
+            ctx2fctx = self.fctxmap[path] # {ctx: fctx}
+            if ctx not in ctx2fctx:
+                continue
+            fctx = ctx2fctx[ctx]
+            content = fctx.data()
+            newcontent = self.fixupmap[path].getfinalcontent(fctx)
+            if content != newcontent:
+                result[fctx.path()] = newcontent
+        return result
+
+    def _movebookmarks(self, tr):
+        repo = self.repo
+        needupdate = [(name, self.replacemap[hsh])
+                      for name, hsh in repo._bookmarks.iteritems()
+                      if hsh in self.replacemap]
+        changes = []
+        for name, hsh in needupdate:
+            if hsh:
+                changes.append((name, hsh))
+                if self.ui.verbose:
+                    self.ui.write(_('moving bookmark %s to %s\n')
+                                  % (name, node.hex(hsh)))
+            else:
+                changes.append((name, None))
+                if self.ui.verbose:
+                    self.ui.write(_('deleting bookmark %s\n') % name)
+        repo._bookmarks.applychanges(repo, tr, changes)
+
+    def _moveworkingdirectoryparent(self):
+        if not self.finalnode:
+            # Find the latest not-{obsoleted,stripped} parent.
+            revs = self.repo.revs('max(::. - %ln)', self.replacemap.keys())
+            ctx = self.repo[revs.first()]
+            self.finalnode = ctx.node()
+        else:
+            ctx = self.repo[self.finalnode]
+
+        dirstate = self.repo.dirstate
+        # dirstate.rebuild invalidates fsmonitorstate, causing "hg status" to
+        # be slow. in absorb's case, no need to invalidate fsmonitorstate.
+        noop = lambda: 0
+        restore = noop
+        if util.safehasattr(dirstate, '_fsmonitorstate'):
+            bak = dirstate._fsmonitorstate.invalidate
+            def restore():
+                dirstate._fsmonitorstate.invalidate = bak
+            dirstate._fsmonitorstate.invalidate = noop
+        try:
+            with dirstate.parentchange():
+                dirstate.rebuild(ctx.node(), ctx.manifest(), self.paths)
+        finally:
+            restore()
+
+    @staticmethod
+    def _willbecomenoop(memworkingcopy, ctx, pctx=None):
+        """({path: content}, ctx, ctx) -> bool. test if a commit will be noop
+
+        if it will become an empty commit (does not change anything, after the
+        memworkingcopy overrides), return True. otherwise return False.
+        """
+        if not pctx:
+            parents = ctx.parents()
+            if len(parents) != 1:
+                return False
+            pctx = parents[0]
+        # ctx changes more files (not a subset of memworkingcopy)
+        if not set(ctx.files()).issubset(set(memworkingcopy.iterkeys())):
+            return False
+        for path, content in memworkingcopy.iteritems():
+            if path not in pctx or path not in ctx:
+                return False
+            fctx = ctx[path]
+            pfctx = pctx[path]
+            if pfctx.flags() != fctx.flags():
+                return False
+            if pfctx.data() != content:
+                return False
+        return True
+
+    def _commitsingle(self, memworkingcopy, ctx, p1=None):
+        """(ctx, {path: content}, node) -> node. make a single commit
+
+        the commit is a clone from ctx, with a (optionally) different p1, and
+        different file contents replaced by memworkingcopy.
+        """
+        parents = p1 and (p1, node.nullid)
+        extra = ctx.extra()
+        if self._useobsolete and self.ui.configbool('absorb', 'addnoise'):
+            extra['absorb_source'] = ctx.hex()
+        mctx = overlaycontext(memworkingcopy, ctx, parents, extra=extra)
+        # preserve phase
+        with mctx.repo().ui.configoverride({
+            ('phases', 'new-commit'): ctx.phase()}):
+            return mctx.commit()
+
+    @util.propertycache
+    def _useobsolete(self):
+        """() -> bool"""
+        return obsolete.isenabled(self.repo, obsolete.createmarkersopt)
+
+    def _obsoleteoldcommits(self):
+        relations = [(self.repo[k], v and (self.repo[v],) or ())
+                     for k, v in self.replacemap.iteritems()]
+        if relations:
+            obsolete.createmarkers(self.repo, relations)
+
+    def _stripoldcommits(self):
+        nodelist = self.replacemap.keys()
+        # make sure we don't strip innocent children
+        revs = self.repo.revs('%ln - (::(heads(%ln::)-%ln))', nodelist,
+                              nodelist, nodelist)
+        tonode = self.repo.changelog.node
+        nodelist = [tonode(r) for r in revs]
+        if nodelist:
+            repair.strip(self.repo.ui, self.repo, nodelist)
+
+def _parsechunk(hunk):
+    """(crecord.uihunk or patch.recordhunk) -> (path, (a1, a2, [bline]))"""
+    if type(hunk) not in (crecord.uihunk, patch.recordhunk):
+        return None, None
+    path = hunk.header.filename()
+    a1 = hunk.fromline + len(hunk.before) - 1
+    # remove before and after context
+    hunk.before = hunk.after = []
+    buf = util.stringio()
+    hunk.write(buf)
+    patchlines = mdiff.splitnewlines(buf.getvalue())
+    # hunk.prettystr() will update hunk.removed
+    a2 = a1 + hunk.removed
+    blines = [l[1:] for l in patchlines[1:] if l[0] != '-']
+    return path, (a1, a2, blines)
+
+def overlaydiffcontext(ctx, chunks):
+    """(ctx, [crecord.uihunk]) -> memctx
+
+    return a memctx with some [1] patches (chunks) applied to ctx.
+    [1]: modifications are handled. renames, mode changes, etc. are ignored.
+    """
+    # sadly the applying-patch logic is hardly reusable, and messy:
+    # 1. the core logic "_applydiff" is too heavy - it writes .rej files, it
+    #    needs a file stream of a patch and will re-parse it, while we have
+    #    structured hunk objects at hand.
+    # 2. a lot of different implementations about "chunk" (patch.hunk,
+    #    patch.recordhunk, crecord.uihunk)
+    # as we only care about applying changes to modified files, no mode
+    # change, no binary diff, and no renames, it's probably okay to
+    # re-invent the logic using much simpler code here.
+    memworkingcopy = {} # {path: content}
+    patchmap = defaultdict(lambda: []) # {path: [(a1, a2, [bline])]}
+    for path, info in map(_parsechunk, chunks):
+        if not path or not info:
+            continue
+        patchmap[path].append(info)
+    for path, patches in patchmap.iteritems():
+        if path not in ctx or not patches:
+            continue
+        patches.sort(reverse=True)
+        lines = mdiff.splitnewlines(ctx[path].data())
+        for a1, a2, blines in patches:
+            lines[a1:a2] = blines
+        memworkingcopy[path] = ''.join(lines)
+    return overlaycontext(memworkingcopy, ctx)
+
+def absorb(ui, repo, stack=None, targetctx=None, pats=None, opts=None):
+    """pick fixup chunks from targetctx, apply them to stack.
+
+    if targetctx is None, the working copy context will be used.
+    if stack is None, the current draft stack will be used.
+    return fixupstate.
+    """
+    if stack is None:
+        limit = ui.configint('absorb', 'maxstacksize')
+        stack = getdraftstack(repo['.'], limit)
+        if limit and len(stack) >= limit:
+            ui.warn(_('absorb: only the recent %d changesets will '
+                      'be analysed\n')
+                    % limit)
+    if not stack:
+        raise error.Abort(_('no changeset to change'))
+    if targetctx is None: # default to working copy
+        targetctx = repo[None]
+    if pats is None:
+        pats = ()
+    if opts is None:
+        opts = {}
+    state = fixupstate(stack, ui=ui, opts=opts)
+    matcher = scmutil.match(targetctx, pats, opts)
+    if opts.get('interactive'):
+        diff = patch.diff(repo, stack[-1].node(), targetctx.node(), matcher)
+        origchunks = patch.parsepatch(diff)
+        chunks = cmdutil.recordfilter(ui, origchunks)[0]
+        targetctx = overlaydiffcontext(stack[-1], chunks)
+    state.diffwith(targetctx, matcher, showchanges=opts.get('print_changes'))
+    if not opts.get('dry_run'):
+        state.apply()
+        if state.commit():
+            state.printchunkstats()
+        elif not ui.quiet:
+            ui.write(_('nothing applied\n'))
+    return state
+
+@command('^absorb|sf',
+         [('p', 'print-changes', None,
+           _('print which changesets are modified by which changes')),
+          ('i', 'interactive', None,
+           _('interactively select which chunks to apply (EXPERIMENTAL)')),
+          ('e', 'edit-lines', None,
+           _('edit what lines belong to which changesets before commit '
+             '(EXPERIMENTAL)')),
+         ] + commands.dryrunopts + commands.walkopts,
+         _('hg absorb [OPTION] [FILE]...'))
+def absorbcmd(ui, repo, *pats, **opts):
+    """incorporate corrections into the stack of draft changesets
+
+    absorb analyzes each change in your working directory and attempts to
+    amend the changed lines into the changesets in your stack that first
+    introduced those lines.
+
+    If absorb cannot find an unambiguous changeset to amend for a change,
+    that change will be left in the working directory, untouched. They can be
+    observed by :hg:`status` or :hg:`diff` afterwards. In other words,
+    absorb does not write to the working directory.
+
+    Changesets outside the revset `::. and not public() and not merge()` will
+    not be changed.
+
+    Changesets that become empty after applying the changes will be deleted.
+
+    If in doubt, run :hg:`absorb -pn` to preview what changesets will
+    be amended by what changed lines, without actually changing anything.
+
+    Returns 0 on success, 1 if all chunks were ignored and nothing amended.
+    """
+    state = absorb(ui, repo, pats=pats, opts=opts)
+    if sum(s[0] for s in state.chunkstats.values()) == 0:
+        return 1
+
+def _wrapamend(flag):
+    """add flag to amend, which will be a shortcut to the absorb command"""
+    if not flag:
+        return
+    amendcmd = extensions.bind(_amendcmd, flag)
+    # the amend command can exist in evolve, or fbamend
+    for extname in ['evolve', 'fbamend', None]:
+        try:
+            if extname is None:
+                cmdtable = commands.table
+            else:
+                ext = extensions.find(extname)
+                cmdtable = ext.cmdtable
+        except (KeyError, AttributeError):
+            continue
+        try:
+            entry = extensions.wrapcommand(cmdtable, 'amend', amendcmd)
+            options = entry[1]
+            msg = _('incorporate corrections into stack. '
+                    'see \'hg help absorb\' for details')
+            options.append(('', flag, None, msg))
+            return
+        except error.UnknownCommand:
+            pass
+
+def _amendcmd(flag, orig, ui, repo, *pats, **opts):
+    if not opts.get(flag):
+        return orig(ui, repo, *pats, **opts)
+    # use absorb
+    for k, v in opts.iteritems(): # check unsupported flags
+        if v and k not in ['interactive', flag]:
+            raise error.Abort(_('--%s does not support --%s')
+                              % (flag, k.replace('_', '-')))
+    state = absorb(ui, repo, pats=pats, opts=opts)
+    # different from the original absorb, tell users what chunks were
+    # ignored and were left. it's because users usually expect "amend" to
+    # take all of their changes and will feel strange otherwise.
+    # the original "absorb" command faces more-advanced users knowing
+    # what's going on and is less verbose.
+    adoptedsum = 0
+    messages = []
+    for path, (adopted, total) in state.chunkstats.iteritems():
+        adoptedsum += adopted
+        if adopted == total:
+            continue
+        reason = _('%d modified chunks were ignored') % (total - adopted)
+        messages.append(('M', 'modified', path, reason))
+    for idx, word, symbol in [(0, 'modified', 'M'), (1, 'added', 'A'),
+                              (2, 'removed', 'R'), (3, 'deleted', '!')]:
+        paths = set(state.status[idx]) - set(state.paths)
+        for path in sorted(paths):
+            if word == 'modified':
+                reason = _('unsupported file type (ex. binary or link)')
+            else:
+                reason = _('%s files were ignored') % word
+            messages.append((symbol, word, path, reason))
+    if messages:
+        ui.write(_('\n# changes not applied and left in '
+                   'working directory:\n'))
+        for symbol, word, path, reason in messages:
+            ui.write(_('# %s %s : %s\n') % (
+                ui.label(symbol, 'status.' + word),
+                ui.label(path, 'status.' + word), reason))
+
+    if adoptedsum == 0:
+        return 1
+
+def extsetup(ui):
+    _wrapamend(ui.config('absorb', 'amendflag'))

tests/test-absorb-edit-lines.t

@@ -0,0 +1,61 b''
+  $ cat >> $HGRCPATH << EOF
+  > [extensions]
+  > absorb=
+  > EOF
+
+  $ hg init repo1
+  $ cd repo1
+
+Make some commits:
+
+  $ for i in 1 2 3; do
+  >   echo $i >> a
+  >   hg commit -A a -m "commit $i" -q
+  > done
+
+absorb --edit-lines will run the editor if filename is provided:
+
+  $ hg absorb --edit-lines
+  nothing applied
+  [1]
+  $ HGEDITOR=cat hg absorb --edit-lines a
+  HG: editing a
+  HG: "y" means the line to the right exists in the changeset to the top
+  HG:
+  HG: /---- 4ec16f85269a commit 1
+  HG: |/--- 5c5f95224a50 commit 2
+  HG: ||/-- 43f0a75bede7 commit 3
+  HG: |||
+      yyy : 1
+       yy : 2
+        y : 3
+  nothing applied
+  [1]
+
+Edit the file using --edit-lines:
+
+  $ cat > editortext << EOF
+  >       y : a
+  >      yy :  b
+  >      y  : c
+  >     yy  : d  
+  >     y y : e
+  >     y   : f
+  >     yyy : g
+  > EOF
+  $ HGEDITOR='cat editortext >' hg absorb -q --edit-lines a
+  $ hg cat -r 0 a
+  d  
+  e
+  f
+  g
+  $ hg cat -r 1 a
+   b
+  c
+  d  
+  g
+  $ hg cat -r 2 a
+  a
+   b
+  e
+  g

tests/test-absorb-filefixupstate.py

@@ -0,0 +1,207 b''
+from __future__ import absolute_import, print_function
+
+import itertools
+
+from hgext import absorb
+
+class simplefctx(object):
+    def __init__(self, content):
+        self.content = content
+
+    def data(self):
+        return self.content
+
+def insertreturns(x):
+    # insert "\n"s after each single char
+    if isinstance(x, str):
+        return ''.join(ch + '\n' for ch in x)
+    else:
+        return map(insertreturns, x)
+
+def removereturns(x):
+    # the revert of "insertreturns"
+    if isinstance(x, str):
+        return x.replace('\n', '')
+    else:
+        return map(removereturns, x)
+
+def assertlistequal(lhs, rhs, decorator=lambda x: x):
+    if lhs != rhs:
+        raise RuntimeError('mismatch:\n actual:   %r\n expected: %r'
+                           % tuple(map(decorator, [lhs, rhs])))
+
+def testfilefixup(oldcontents, workingcopy, expectedcontents, fixups=None):
+    """([str], str, [str], [(rev, a1, a2, b1, b2)]?) -> None
+
+    workingcopy is a string, of which every character denotes a single line.
+
+    oldcontents, expectedcontents are lists of strings, every character of
+    every string denots a single line.
+
+    if fixups is not None, it's the expected fixups list and will be checked.
+    """
+    expectedcontents = insertreturns(expectedcontents)
+    oldcontents = insertreturns(oldcontents)
+    workingcopy = insertreturns(workingcopy)
+    state = absorb.filefixupstate(map(simplefctx, oldcontents))
+    state.diffwith(simplefctx(workingcopy))
+    if fixups is not None:
+        assertlistequal(state.fixups, fixups)
+    state.apply()
+    assertlistequal(state.finalcontents, expectedcontents, removereturns)
+
+def buildcontents(linesrevs):
+    # linesrevs: [(linecontent : str, revs : [int])]
+    revs = set(itertools.chain(*[revs for line, revs in linesrevs]))
+    return [''] + [
+        ''.join([l for l, rs in linesrevs if r in rs])
+        for r in sorted(revs)
+    ]
+
+# input case 0: one single commit
+case0 = ['', '11']
+
+# replace a single chunk
+testfilefixup(case0, '', ['', ''])
+testfilefixup(case0, '2', ['', '2'])
+testfilefixup(case0, '22', ['', '22'])
+testfilefixup(case0, '222', ['', '222'])
+
+# input case 1: 3 lines, each commit adds one line
+case1 = buildcontents([
+    ('1', [1, 2, 3]),
+    ('2', [   2, 3]),
+    ('3', [      3]),
+])
+
+# 1:1 line mapping
+testfilefixup(case1, '123', case1)
+testfilefixup(case1, '12c', ['', '1', '12', '12c'])
+testfilefixup(case1, '1b3', ['', '1', '1b', '1b3'])
+testfilefixup(case1, '1bc', ['', '1', '1b', '1bc'])
+testfilefixup(case1, 'a23', ['', 'a', 'a2', 'a23'])
+testfilefixup(case1, 'a2c', ['', 'a', 'a2', 'a2c'])
+testfilefixup(case1, 'ab3', ['', 'a', 'ab', 'ab3'])
+testfilefixup(case1, 'abc', ['', 'a', 'ab', 'abc'])
+
+# non 1:1 edits
+testfilefixup(case1, 'abcd', case1)
+testfilefixup(case1, 'ab', case1)
+
+# deletion
+testfilefixup(case1, '',   ['', '', '', ''])
+testfilefixup(case1, '1',  ['', '1', '1', '1'])
+testfilefixup(case1, '2',  ['', '', '2', '2'])
+testfilefixup(case1, '3',  ['', '', '', '3'])
+testfilefixup(case1, '13', ['', '1', '1', '13'])
+
+# replaces
+testfilefixup(case1, '1bb3', ['', '1', '1bb', '1bb3'])
+
+# (confusing) replaces
+testfilefixup(case1, '1bbb', case1)
+testfilefixup(case1, 'bbbb', case1)
+testfilefixup(case1, 'bbb3', case1)
+testfilefixup(case1, '1b', case1)
+testfilefixup(case1, 'bb', case1)
+testfilefixup(case1, 'b3', case1)
+
+# insertions at the beginning and the end
+testfilefixup(case1, '123c', ['', '1', '12', '123c'])
+testfilefixup(case1, 'a123', ['', 'a1', 'a12', 'a123'])
+
+# (confusing) insertions
+testfilefixup(case1, '1a23', case1)
+testfilefixup(case1, '12b3', case1)
+
+# input case 2: delete in the middle
+case2 = buildcontents([
+    ('11', [1, 2]),
+    ('22', [1   ]),
+    ('33', [1, 2]),
+])
+
+# deletion (optimize code should make it 2 chunks)
+testfilefixup(case2, '', ['', '22', ''],
+              fixups=[(4, 0, 2, 0, 0), (4, 2, 4, 0, 0)])
+
+# 1:1 line mapping
+testfilefixup(case2, 'aaaa', ['', 'aa22aa', 'aaaa'])
+
+# non 1:1 edits
+# note: unlike case0, the chunk is not "continuous" and no edit allowed
+testfilefixup(case2, 'aaa', case2)
+
+# input case 3: rev 3 reverts rev 2
+case3 = buildcontents([
+    ('1', [1, 2, 3]),
+    ('2', [   2   ]),
+    ('3', [1, 2, 3]),
+])
+
+# 1:1 line mapping
+testfilefixup(case3, '13', case3)
+testfilefixup(case3, '1b', ['', '1b', '12b', '1b'])
+testfilefixup(case3, 'a3', ['', 'a3', 'a23', 'a3'])
+testfilefixup(case3, 'ab', ['', 'ab', 'a2b', 'ab'])
+
+# non 1:1 edits
+testfilefixup(case3, 'a', case3)
+testfilefixup(case3, 'abc', case3)
+
+# deletion
+testfilefixup(case3, '', ['', '', '2', ''])
+
+# insertion
+testfilefixup(case3, 'a13c', ['', 'a13c', 'a123c', 'a13c'])
+
+# input case 4: a slightly complex case
+case4 = buildcontents([
+    ('1', [1, 2, 3]),
+    ('2', [   2, 3]),
+    ('3', [1, 2,  ]),
+    ('4', [1,    3]),
+    ('5', [      3]),
+    ('6', [   2, 3]),
+    ('7', [   2   ]),
+    ('8', [   2, 3]),
+    ('9', [      3]),
+])
+
+testfilefixup(case4, '1245689', case4)
+testfilefixup(case4, '1a2456bbb', case4)
+testfilefixup(case4, '1abc5689', case4)
+testfilefixup(case4, '1ab5689', ['', '134', '1a3678', '1ab5689'])
+testfilefixup(case4, 'aa2bcd8ee', ['', 'aa34', 'aa23d78', 'aa2bcd8ee'])
+testfilefixup(case4, 'aa2bcdd8ee',['', 'aa34', 'aa23678', 'aa24568ee'])
+testfilefixup(case4, 'aaaaaa', case4)
+testfilefixup(case4, 'aa258b', ['', 'aa34', 'aa2378', 'aa258b'])
+testfilefixup(case4, '25bb', ['', '34', '23678', '25689'])
+testfilefixup(case4, '27', ['', '34', '23678', '245689'])
+testfilefixup(case4, '28', ['', '34', '2378', '28'])
+testfilefixup(case4, '', ['', '34', '37', ''])
+
+# input case 5: replace a small chunk which is near a deleted line
+case5 = buildcontents([
+    ('12', [1, 2]),
+    ('3',  [1]),
+    ('4',  [1, 2]),
+])
+
+testfilefixup(case5, '1cd4', ['', '1cd34', '1cd4'])
+
+# input case 6: base "changeset" is immutable
+case6 = ['1357', '0125678']
+
+testfilefixup(case6, '0125678', case6)
+testfilefixup(case6, '0a25678', case6)
+testfilefixup(case6, '0a256b8', case6)
+testfilefixup(case6, 'abcdefg', ['1357', 'a1c5e7g'])
+testfilefixup(case6, 'abcdef', case6)
+testfilefixup(case6, '', ['1357', '157'])
+testfilefixup(case6, '0123456789', ['1357', '0123456789'])
+
+# input case 7: change an empty file
+case7 = ['']
+
+testfilefixup(case7, '1', case7)

tests/test-absorb-phase.t

@@ -0,0 +1,30 b''
+  $ cat >> $HGRCPATH << EOF
+  > [extensions]
+  > absorb=
+  > drawdag=$RUNTESTDIR/drawdag.py
+  > EOF
+
+  $ hg init
+  $ hg debugdrawdag <<'EOS'
+  > C
+  > |
+  > B
+  > |
+  > A
+  > EOS
+
+  $ hg phase -r A --public -q
+  $ hg phase -r C --secret --force -q
+
+  $ hg update C -q
+  $ printf B1 > B
+
+  $ hg absorb -q
+
+  $ hg log -G -T '{desc} {phase}'
+  @  C secret
+  |
+  o  B draft
+  |
+  o  A public
+  

tests/test-absorb-rename.t

@@ -0,0 +1,359 b''
+  $ cat >> $HGRCPATH << EOF
+  > [diff]
+  > git=1
+  > [extensions]
+  > absorb=
+  > EOF
+
+  $ sedi() { # workaround check-code
+  > pattern="$1"
+  > shift
+  > for i in "$@"; do
+  >     sed "$pattern" "$i" > "$i".tmp
+  >     mv "$i".tmp "$i"
+  > done
+  > }
+
+rename a to b, then b to a
+
+  $ hg init repo1
+  $ cd repo1
+
+  $ echo 1 > a
+  $ hg ci -A a -m 1
+  $ hg mv a b
+  $ echo 2 >> b
+  $ hg ci -m 2
+  $ hg mv b a
+  $ echo 3 >> a
+  $ hg ci -m 3
+
+  $ hg annotate -ncf a
+  0 eff892de26ec a: 1
+  1 bf56e1f4f857 b: 2
+  2 0b888b00216c a: 3
+
+  $ sedi 's/$/a/' a
+  $ hg absorb -pq
+  showing changes for a
+          @@ -0,3 +0,3 @@
+  eff892d -1
+  bf56e1f -2
+  0b888b0 -3
+  eff892d +1a
+  bf56e1f +2a
+  0b888b0 +3a
+
+  $ hg status
+
+  $ hg annotate -ncf a
+  0 5d1c5620e6f2 a: 1a
+  1 9a14ffe67ae9 b: 2a
+  2 9191d121a268 a: 3a
+
+when the first changeset is public
+
+  $ hg phase --public -r 0
+
+  $ sedi 's/a/A/' a
+
+  $ hg absorb -pq
+  showing changes for a
+          @@ -0,3 +0,3 @@
+          -1a
+  9a14ffe -2a
+  9191d12 -3a
+          +1A
+  9a14ffe +2A
+  9191d12 +3A
+
+  $ hg diff
+  diff --git a/a b/a
+  --- a/a
+  +++ b/a
+  @@ -1,3 +1,3 @@
+  -1a
+  +1A
+   2A
+   3A
+
+copy a to b
+
+  $ cd ..
+  $ hg init repo2
+  $ cd repo2
+
+  $ echo 1 > a
+  $ hg ci -A a -m 1
+  $ hg cp a b
+  $ echo 2 >> b
+  $ hg ci -m 2
+
+  $ hg log -T '{rev}:{node|short} {desc}\n'
+  1:17b72129ab68 2
+  0:eff892de26ec 1
+
+  $ sedi 's/$/a/' a
+  $ sedi 's/$/b/' b
+
+  $ hg absorb -pq
+  showing changes for a
+          @@ -0,1 +0,1 @@
+  eff892d -1
+  eff892d +1a
+  showing changes for b
+          @@ -0,2 +0,2 @@
+          -1
+  17b7212 -2
+          +1b
+  17b7212 +2b
+
+  $ hg diff
+  diff --git a/b b/b
+  --- a/b
+  +++ b/b
+  @@ -1,2 +1,2 @@
+  -1
+  +1b
+   2b
+
+copy b to a
+
+  $ cd ..
+  $ hg init repo3
+  $ cd repo3
+
+  $ echo 1 > b
+  $ hg ci -A b -m 1
+  $ hg cp b a
+  $ echo 2 >> a
+  $ hg ci -m 2
+
+  $ hg log -T '{rev}:{node|short} {desc}\n'
+  1:e62c256d8b24 2
+  0:55105f940d5c 1
+
+  $ sedi 's/$/a/' a
+  $ sedi 's/$/a/' b
+
+  $ hg absorb -pq
+  showing changes for a
+          @@ -0,2 +0,2 @@
+          -1
+  e62c256 -2
+          +1a
+  e62c256 +2a
+  showing changes for b
+          @@ -0,1 +0,1 @@
+  55105f9 -1
+  55105f9 +1a
+
+  $ hg diff
+  diff --git a/a b/a
+  --- a/a
+  +++ b/a
+  @@ -1,2 +1,2 @@
+  -1
+  +1a
+   2a
+
+"move" b to both a and c, follow a - sorted alphabetically
+
+  $ cd ..
+  $ hg init repo4
+  $ cd repo4
+
+  $ echo 1 > b
+  $ hg ci -A b -m 1
+  $ hg cp b a
+  $ hg cp b c
+  $ hg rm b
+  $ echo 2 >> a
+  $ echo 3 >> c
+  $ hg commit -m cp
+
+  $ hg log -T '{rev}:{node|short} {desc}\n'
+  1:366daad8e679 cp
+  0:55105f940d5c 1
+
+  $ sedi 's/$/a/' a
+  $ sedi 's/$/c/' c
+
+  $ hg absorb -pq
+  showing changes for a
+          @@ -0,2 +0,2 @@
+  55105f9 -1
+  366daad -2
+  55105f9 +1a
+  366daad +2a
+  showing changes for c
+          @@ -0,2 +0,2 @@
+          -1
+  366daad -3
+          +1c
+  366daad +3c
+
+  $ hg log -G -p -T '{rev}:{node|short} {desc}\n'
+  @  1:70606019f91b cp
+  |  diff --git a/b b/a
+  |  rename from b
+  |  rename to a
+  |  --- a/b
+  |  +++ b/a
+  |  @@ -1,1 +1,2 @@
+  |   1a
+  |  +2a
+  |  diff --git a/b b/c
+  |  copy from b
+  |  copy to c
+  |  --- a/b
+  |  +++ b/c
+  |  @@ -1,1 +1,2 @@
+  |  -1a
+  |  +1
+  |  +3c
+  |
+  o  0:bfb67c3539c1 1
+     diff --git a/b b/b
+     new file mode 100644
+     --- /dev/null
+     +++ b/b
+     @@ -0,0 +1,1 @@
+     +1a
+  
+run absorb again would apply the change to c
+
+  $ hg absorb -pq
+  showing changes for c
+          @@ -0,1 +0,1 @@
+  7060601 -1
+  7060601 +1c
+
+  $ hg log -G -p -T '{rev}:{node|short} {desc}\n'
+  @  1:8bd536cce368 cp
+  |  diff --git a/b b/a
+  |  rename from b
+  |  rename to a
+  |  --- a/b
+  |  +++ b/a
+  |  @@ -1,1 +1,2 @@
+  |   1a
+  |  +2a
+  |  diff --git a/b b/c
+  |  copy from b
+  |  copy to c
+  |  --- a/b
+  |  +++ b/c
+  |  @@ -1,1 +1,2 @@
+  |  -1a
+  |  +1c
+  |  +3c
+  |
+  o  0:bfb67c3539c1 1
+     diff --git a/b b/b
+     new file mode 100644
+     --- /dev/null
+     +++ b/b
+     @@ -0,0 +1,1 @@
+     +1a
+  
+"move" b to a, c and d, follow d if a gets renamed to e, and c is deleted
+
+  $ cd ..
+  $ hg init repo5
+  $ cd repo5
+
+  $ echo 1 > b
+  $ hg ci -A b -m 1
+  $ hg cp b a
+  $ hg cp b c
+  $ hg cp b d
+  $ hg rm b
+  $ echo 2 >> a
+  $ echo 3 >> c
+  $ echo 4 >> d
+  $ hg commit -m cp
+  $ hg mv a e
+  $ hg rm c
+  $ hg commit -m mv
+
+  $ hg log -T '{rev}:{node|short} {desc}\n'
+  2:49911557c471 mv
+  1:7bc3d43ede83 cp
+  0:55105f940d5c 1
+
+  $ sedi 's/$/e/' e
+  $ sedi 's/$/d/' d
+
+  $ hg absorb -pq
+  showing changes for d
+          @@ -0,2 +0,2 @@
+  55105f9 -1
+  7bc3d43 -4
+  55105f9 +1d
+  7bc3d43 +4d
+  showing changes for e
+          @@ -0,2 +0,2 @@
+          -1
+  7bc3d43 -2
+          +1e
+  7bc3d43 +2e
+
+  $ hg diff
+  diff --git a/e b/e
+  --- a/e
+  +++ b/e
+  @@ -1,2 +1,2 @@
+  -1
+  +1e
+   2e
+
+  $ hg log -G -p -T '{rev}:{node|short} {desc}\n'
+  @  2:34be9b0c786e mv
+  |  diff --git a/c b/c
+  |  deleted file mode 100644
+  |  --- a/c
+  |  +++ /dev/null
+  |  @@ -1,2 +0,0 @@
+  |  -1
+  |  -3
+  |  diff --git a/a b/e
+  |  rename from a
+  |  rename to e
+  |
+  o  1:13e56db5948d cp
+  |  diff --git a/b b/a
+  |  rename from b
+  |  rename to a
+  |  --- a/b
+  |  +++ b/a
+  |  @@ -1,1 +1,2 @@
+  |  -1d
+  |  +1
+  |  +2e
+  |  diff --git a/b b/c
+  |  copy from b
+  |  copy to c
+  |  --- a/b
+  |  +++ b/c
+  |  @@ -1,1 +1,2 @@
+  |  -1d
+  |  +1
+  |  +3
+  |  diff --git a/b b/d
+  |  copy from b
+  |  copy to d
+  |  --- a/b
+  |  +++ b/d
+  |  @@ -1,1 +1,2 @@
+  |   1d
+  |  +4d
+  |
+  o  0:0037613a5dc6 1
+     diff --git a/b b/b
+     new file mode 100644
+     --- /dev/null
+     +++ b/b
+     @@ -0,0 +1,1 @@
+     +1d
+  

tests/test-absorb-strip.t

@@ -0,0 +1,45 b''
+Do not strip innocent children. See https://bitbucket.org/facebook/hg-experimental/issues/6/hg-absorb-merges-diverged-commits
+
+  $ cat >> $HGRCPATH << EOF
+  > [extensions]
+  > absorb=
+  > drawdag=$RUNTESTDIR/drawdag.py
+  > EOF
+
+  $ hg init
+  $ hg debugdrawdag << EOF
+  > E
+  > |
+  > D F
+  > |/
+  > C
+  > |
+  > B
+  > |
+  > A
+  > EOF
+
+  $ hg up E -q
+  $ echo 1 >> B
+  $ echo 2 >> D
+  $ hg absorb
+  saved backup bundle to * (glob)
+  2 of 2 chunk(s) applied
+
+  $ hg log -G -T '{desc}'
+  @  E
+  |
+  o  D
+  |
+  o  C
+  |
+  o  B
+  |
+  | o  F
+  | |
+  | o  C
+  | |
+  | o  B
+  |/
+  o  A
+  

tests/test-absorb.t

@@ -0,0 +1,478 b''
+  $ cat >> $HGRCPATH << EOF
+  > [extensions]
+  > absorb=
+  > EOF
+
+  $ sedi() { # workaround check-code
+  > pattern="$1"
+  > shift
+  > for i in "$@"; do
+  >     sed "$pattern" "$i" > "$i".tmp
+  >     mv "$i".tmp "$i"
+  > done
+  > }
+
+  $ hg init repo1
+  $ cd repo1
+
+Do not crash with empty repo:
+
+  $ hg absorb
+  abort: no changeset to change
+  [255]
+
+Make some commits:
+
+  $ for i in 1 2 3 4 5; do
+  >   echo $i >> a
+  >   hg commit -A a -m "commit $i" -q
+  > done
+
+  $ hg annotate a
+  0: 1
+  1: 2
+  2: 3
+  3: 4
+  4: 5
+
+Change a few lines:
+
+  $ cat > a <<EOF
+  > 1a
+  > 2b
+  > 3
+  > 4d
+  > 5e
+  > EOF
+
+Preview absorb changes:
+
+  $ hg absorb --print-changes --dry-run
+  showing changes for a
+          @@ -0,2 +0,2 @@
+  4ec16f8 -1
+  5c5f952 -2
+  4ec16f8 +1a
+  5c5f952 +2b
+          @@ -3,2 +3,2 @@
+  ad8b8b7 -4
+  4f55fa6 -5
+  ad8b8b7 +4d
+  4f55fa6 +5e
+
+Run absorb:
+
+  $ hg absorb
+  saved backup bundle to * (glob)
+  2 of 2 chunk(s) applied
+  $ hg annotate a
+  0: 1a
+  1: 2b
+  2: 3
+  3: 4d
+  4: 5e
+
+Delete a few lines and related commits will be removed if they will be empty:
+
+  $ cat > a <<EOF
+  > 2b
+  > 4d
+  > EOF
+  $ hg absorb
+  saved backup bundle to * (glob)
+  3 of 3 chunk(s) applied
+  $ hg annotate a
+  1: 2b
+  2: 4d
+  $ hg log -T '{rev} {desc}\n' -Gp
+  @  2 commit 4
+  |  diff -r 1cae118c7ed8 -r 58a62bade1c6 a
+  |  --- a/a	Thu Jan 01 00:00:00 1970 +0000
+  |  +++ b/a	Thu Jan 01 00:00:00 1970 +0000
+  |  @@ -1,1 +1,2 @@
+  |   2b
+  |  +4d
+  |
+  o  1 commit 2
+  |  diff -r 84add69aeac0 -r 1cae118c7ed8 a
+  |  --- a/a	Thu Jan 01 00:00:00 1970 +0000
+  |  +++ b/a	Thu Jan 01 00:00:00 1970 +0000
+  |  @@ -0,0 +1,1 @@
+  |  +2b
+  |
+  o  0 commit 1
+  
+
+Non 1:1 map changes will be ignored:
+
+  $ echo 1 > a
+  $ hg absorb
+  nothing applied
+  [1]
+
+Insertaions:
+
+  $ cat > a << EOF
+  > insert before 2b
+  > 2b
+  > 4d
+  > insert aftert 4d
+  > EOF
+  $ hg absorb -q
+  $ hg status
+  $ hg annotate a
+  1: insert before 2b
+  1: 2b
+  2: 4d
+  2: insert aftert 4d
+
+Bookmarks are moved:
+
+  $ hg bookmark -r 1 b1
+  $ hg bookmark -r 2 b2
+  $ hg bookmark ba
+  $ hg bookmarks
+     b1                        1:b35060a57a50
+     b2                        2:946e4bc87915
+   * ba                        2:946e4bc87915
+  $ sedi 's/insert/INSERT/' a
+  $ hg absorb -q
+  $ hg status
+  $ hg bookmarks
+     b1                        1:a4183e9b3d31
+     b2                        2:c9b20c925790
+   * ba                        2:c9b20c925790
+
+Non-mofified files are ignored:
+
+  $ touch b
+  $ hg commit -A b -m b
+  $ touch c
+  $ hg add c
+  $ hg rm b
+  $ hg absorb
+  nothing applied
+  [1]
+  $ sedi 's/INSERT/Insert/' a
+  $ hg absorb
+  saved backup bundle to * (glob)
+  2 of 2 chunk(s) applied
+  $ hg status
+  A c
+  R b
+
+Public commits will not be changed:
+
+  $ hg phase -p 1
+  $ sedi 's/Insert/insert/' a
+  $ hg absorb -pn
+  showing changes for a
+          @@ -0,1 +0,1 @@
+          -Insert before 2b
+          +insert before 2b
+          @@ -3,1 +3,1 @@
+  85b4e0e -Insert aftert 4d
+  85b4e0e +insert aftert 4d
+  $ hg absorb
+  saved backup bundle to * (glob)
+  1 of 2 chunk(s) applied
+  $ hg diff -U 0
+  diff -r 1c8eadede62a a
+  --- a/a	Thu Jan 01 00:00:00 1970 +0000
+  +++ b/a	* (glob)
+  @@ -1,1 +1,1 @@
+  -Insert before 2b
+  +insert before 2b
+  $ hg annotate a
+  1: Insert before 2b
+  1: 2b
+  2: 4d
+  2: insert aftert 4d
+
+Make working copy clean:
+
+  $ hg revert -q -C a b
+  $ hg forget c
+  $ rm c
+  $ hg status
+
+Merge commit will not be changed:
+
+  $ echo 1 > m1
+  $ hg commit -A m1 -m m1
+  $ hg bookmark -q -i m1
+  $ hg update -q '.^'
+  $ echo 2 > m2
+  $ hg commit -q -A m2 -m m2
+  $ hg merge -q m1
+  $ hg commit -m merge
+  $ hg bookmark -d m1
+  $ hg log -G -T '{rev} {desc} {phase}\n'
+  @    6 merge draft
+  |\
+  | o  5 m2 draft
+  | |
+  o |  4 m1 draft
+  |/
+  o  3 b draft
+  |
+  o  2 commit 4 draft
+  |
+  o  1 commit 2 public
+  |
+  o  0 commit 1 public
+  
+  $ echo 2 >> m1
+  $ echo 2 >> m2
+  $ hg absorb
+  abort: no changeset to change
+  [255]
+  $ hg revert -q -C m1 m2
+
+Use a new repo:
+
+  $ cd ..
+  $ hg init repo2
+  $ cd repo2
+
+Make some commits to multiple files:
+
+  $ for f in a b; do
+  >   for i in 1 2; do
+  >     echo $f line $i >> $f
+  >     hg commit -A $f -m "commit $f $i" -q
+  >   done
+  > done
+
+Use pattern to select files to be fixed up:
+
+  $ sedi 's/line/Line/' a b
+  $ hg status
+  M a
+  M b
+  $ hg absorb a
+  saved backup bundle to * (glob)
+  1 of 1 chunk(s) applied
+  $ hg status
+  M b
+  $ hg absorb --exclude b
+  nothing applied
+  [1]
+  $ hg absorb b
+  saved backup bundle to * (glob)
+  1 of 1 chunk(s) applied
+  $ hg status
+  $ cat a b
+  a Line 1
+  a Line 2
+  b Line 1
+  b Line 2
+
+Test config option absorb.maxstacksize:
+
+  $ sedi 's/Line/line/' a b
+  $ hg log -T '{rev}:{node} {desc}\n'
+  3:712d16a8f445834e36145408eabc1d29df05ec09 commit b 2
+  2:74cfa6294160149d60adbf7582b99ce37a4597ec commit b 1
+  1:28f10dcf96158f84985358a2e5d5b3505ca69c22 commit a 2
+  0:f9a81da8dc53380ed91902e5b82c1b36255a4bd0 commit a 1
+  $ hg --config absorb.maxstacksize=1 absorb -pn
+  absorb: only the recent 1 changesets will be analysed
+  showing changes for a
+          @@ -0,2 +0,2 @@
+          -a Line 1
+          -a Line 2
+          +a line 1
+          +a line 2
+  showing changes for b
+          @@ -0,2 +0,2 @@
+          -b Line 1
+  712d16a -b Line 2
+          +b line 1
+  712d16a +b line 2
+
+Test obsolete markers creation:
+
+  $ cat >> $HGRCPATH << EOF
+  > [experimental]
+  > evolution=createmarkers
+  > [absorb]
+  > addnoise=1
+  > EOF
+
+  $ hg --config absorb.maxstacksize=3 sf
+  absorb: only the recent 3 changesets will be analysed
+  2 of 2 chunk(s) applied
+  $ hg log -T '{rev}:{node|short} {desc} {get(extras, "absorb_source")}\n'
+  6:3dfde4199b46 commit b 2 712d16a8f445834e36145408eabc1d29df05ec09
+  5:99cfab7da5ff commit b 1 74cfa6294160149d60adbf7582b99ce37a4597ec
+  4:fec2b3bd9e08 commit a 2 28f10dcf96158f84985358a2e5d5b3505ca69c22
+  0:f9a81da8dc53 commit a 1 
+  $ hg absorb
+  1 of 1 chunk(s) applied
+  $ hg log -T '{rev}:{node|short} {desc} {get(extras, "absorb_source")}\n'
+  10:e1c8c1e030a4 commit b 2 3dfde4199b4610ea6e3c6fa9f5bdad8939d69524
+  9:816c30955758 commit b 1 99cfab7da5ffdaf3b9fc6643b14333e194d87f46
+  8:5867d584106b commit a 2 fec2b3bd9e0834b7cb6a564348a0058171aed811
+  7:8c76602baf10 commit a 1 f9a81da8dc53380ed91902e5b82c1b36255a4bd0
+
+Test config option absorb.amendflags and running as a sub command of amend:
+
+  $ cat >> $TESTTMP/dummyamend.py << EOF
+  > from mercurial import commands, registrar
+  > cmdtable = {}
+  > command = registrar.command(cmdtable)
+  > @command('amend', [], '')
+  > def amend(ui, repo, *pats, **opts):
+  >     return 3
+  > EOF
+  $ cat >> $HGRCPATH << EOF
+  > [extensions]
+  > fbamend=$TESTTMP/dummyamend.py
+  > [absorb]
+  > amendflag = correlated
+  > EOF
+
+  $ hg amend -h
+  hg amend
+  
+  (no help text available)
+  
+  options:
+  
+    --correlated incorporate corrections into stack. see 'hg help absorb' for
+                 details
+  
+  (some details hidden, use --verbose to show complete help)
+
+  $ $PYTHON -c 'print("".join(map(chr, range(0,3))))' > c
+  $ hg commit -A c -m 'c is a binary file'
+  $ echo c >> c
+  $ sedi $'2i\\\nINS\n' b
+  $ echo END >> b
+  $ hg rm a
+  $ hg amend --correlated
+  1 of 2 chunk(s) applied
+  
+  # changes not applied and left in working directory:
+  # M b : 1 modified chunks were ignored
+  # M c : unsupported file type (ex. binary or link)
+  # R a : removed files were ignored
+
+Executable files:
+
+  $ cat >> $HGRCPATH << EOF
+  > [diff]
+  > git=True
+  > EOF
+  $ cd ..
+  $ hg init repo3
+  $ cd repo3
+  $ echo > foo.py
+  $ chmod +x foo.py
+  $ hg add foo.py
+  $ hg commit -mfoo
+
+  $ echo bla > foo.py
+  $ hg absorb --dry-run --print-changes
+  showing changes for foo.py
+          @@ -0,1 +0,1 @@
+  99b4ae7 -
+  99b4ae7 +bla
+  $ hg absorb
+  1 of 1 chunk(s) applied
+  $ hg diff -c .
+  diff --git a/foo.py b/foo.py
+  new file mode 100755
+  --- /dev/null
+  +++ b/foo.py
+  @@ -0,0 +1,1 @@
+  +bla
+  $ hg diff
+
+Remove lines may delete changesets:
+
+  $ cd ..
+  $ hg init repo4
+  $ cd repo4
+  $ cat > a <<EOF
+  > 1
+  > 2
+  > EOF
+  $ hg commit -m a12 -A a
+  $ cat > b <<EOF
+  > 1
+  > 2
+  > EOF
+  $ hg commit -m b12 -A b
+  $ echo 3 >> b
+  $ hg commit -m b3
+  $ echo 4 >> b
+  $ hg commit -m b4
+  $ echo 1 > b
+  $ echo 3 >> a
+  $ hg absorb -pn
+  showing changes for a
+          @@ -2,0 +2,1 @@
+  bfafb49 +3
+  showing changes for b
+          @@ -1,3 +1,0 @@
+  1154859 -2
+  30970db -3
+  a393a58 -4
+  $ hg absorb -v | grep became
+  bfafb49242db: 1 file(s) changed, became 1a2de97fc652
+  115485984805: 2 file(s) changed, became 0c930dfab74c
+  30970dbf7b40: became empty and was dropped
+  a393a58b9a85: became empty and was dropped
+  $ hg log -T '{rev} {desc}\n' -Gp
+  @  5 b12
+  |  diff --git a/b b/b
+  |  new file mode 100644
+  |  --- /dev/null
+  |  +++ b/b
+  |  @@ -0,0 +1,1 @@
+  |  +1
+  |
+  o  4 a12
+     diff --git a/a b/a
+     new file mode 100644
+     --- /dev/null
+     +++ b/a
+     @@ -0,0 +1,3 @@
+     +1
+     +2
+     +3
+  
+
+Use revert to make the current change and its parent disappear.
+This should move us to the non-obsolete ancestor.
+
+  $ cd ..
+  $ hg init repo5
+  $ cd repo5
+  $ cat > a <<EOF
+  > 1
+  > 2
+  > EOF
+  $ hg commit -m a12 -A a
+  $ hg id
+  bfafb49242db tip
+  $ echo 3 >> a
+  $ hg commit -m a123 a
+  $ echo 4 >> a
+  $ hg commit -m a1234 a
+  $ hg id
+  82dbe7fd19f0 tip
+  $ hg revert -r 0 a
+  $ hg absorb -pn
+  showing changes for a
+          @@ -2,2 +2,0 @@
+  f1c23dd -3
+  82dbe7f -4
+  $ hg absorb --verbose
+  f1c23dd5d08d: became empty and was dropped
+  82dbe7fd19f0: became empty and was dropped
+  a: 1 of 1 chunk(s) applied
+  $ hg id
+  bfafb49242db tip