This is an archive of the discontinued Mercurial Phabricator instance.

Differential D12418

copies: add config to preserve old copies (issue5457)
Needs ReviewPublic

Authored by martinvonz on Mar 30 2022, 12:18 AM.

Details

Reviewers
None
Group Reviewers
hg-reviewers
Summary

The scenario in issue 5457 gets reported every now and then by our
users (Google developers). This patch adds a fix for it by preserving
existing copies even if the source file no longer exists. That's a bit
of hack, but it's effective and much simpler than the other
alternatives we've considered (such as storing the copy information in
obsmarkers, which would mean that copy tracing would have to follow
both graphs).

The fix only works when copies are stored in changeset extras or
sidedata because the filelog record needs to have the old file nodeid,
which doesn't exist in the parent manifest.

Diff Detail

Repository
rHG Mercurial
Branch
default
Lint
No Linters Available
Unit
No Unit Test Coverage

Event Timeline

martinvonz created this revision.Mar 30 2022, 12:18 AM
marmoute added a subscriber: marmoute.Apr 2 2022, 1:29 PM

What happens to the old (now invalid) value after the rebase ?

martinvonz added a comment.Apr 4 2022, 12:56 PM
In D12418#190140, @marmoute wrote:

What happens to the old (now invalid) value after the rebase ?

Which "value" do you mean? The copy information is stored in the amended commit, so it's not affected by the rebase. It will remain there after commits are rebased on top as usual.

Alphare added a subscriber: Alphare.Apr 6 2022, 6:28 AM

As a knee-jerk reaction to the commit message only, I'd say that I am *very* uncomfortable essentially keeping broken data around. I will need to think about this when I have more time than today, I wanted to say something soon so you wouldn't think this went overlooked.

Revision Contents
Changeset List

Diff 32732

mercurial/commit.py

# behavior in this circumstance. # behavior in this circumstance.
if cnode: if cnode:
repo.ui.debug(b" %s: copy %s:%s\n" % (fname, cfname, hex(cnode))) repo.ui.debug(b" %s: copy %s:%s\n" % (fname, cfname, hex(cnode)))
if includecopymeta: if includecopymeta:
meta[b"copy"] = cfname meta[b"copy"] = cfname
meta[b"copyrev"] = hex(cnode) meta[b"copyrev"] = hex(cnode)
fparent1, fparent2 = repo.nullid, newfparent fparent1, fparent2 = repo.nullid, newfparent
else: elif not repo.ui.configbool(b'experimental', b'copies.keep-old'):
repo.ui.warn( repo.ui.warn(
_( _(
b"warning: can't find ancestor for '%s' " b"warning: can't find ancestor for '%s' "
b"copied from '%s'!\n" b"copied from '%s'!\n"
) )
% (fname, cfname) % (fname, cfname)
) )

mercurial/configitems.py

) )
coreconfigitem( coreconfigitem(
b'experimental', b'experimental',
b'copies.write-to', b'copies.write-to',
default=b'filelog-only', default=b'filelog-only',
) )
coreconfigitem( coreconfigitem(
b'experimental', b'experimental',
b'copies.keep-old',
default=False,
)
coreconfigitem(
b'experimental',
b'crecordtest', b'crecordtest',
default=None, default=None,
) )
coreconfigitem( coreconfigitem(
b'experimental', b'experimental',
b'directaccess', b'directaccess',
default=False, default=False,
) )

mercurial/copies.py

# 5 - x y - # 5 - x y -
# 6 x x y x->y # 6 x x y x->y
# #
# _chain() takes care of chaining the copies in 'a' and 'b', but it # _chain() takes care of chaining the copies in 'a' and 'b', but it
# cannot tell the difference between cases 1 and 2, between 3 and 4, or # cannot tell the difference between cases 1 and 2, between 3 and 4, or
# between 5 and 6, so it includes all cases in its result. # between 5 and 6, so it includes all cases in its result.
# Cases 1, 3, and 5 are then removed by _filter(). # Cases 1, 3, and 5 are then removed by _filter().
keep_old = src.repo().ui.configbool(b'experimental', b'copies.keep-old')
for k, v in list(t.items()): for k, v in list(t.items()):
if k == v: # case 3 if k == v: # case 3
del t[k] del t[k]
elif v not in src: # case 5 elif not keep_old and v not in src: # case 5
# remove copies from files that didn't exist # remove copies from files that didn't exist
del t[k] del t[k]
elif k not in dst: # case 1 elif k not in dst: # case 1
# remove copies to files that were then removed # remove copies to files that were then removed
del t[k] del t[k]
def _chain(prefix, suffix): def _chain(prefix, suffix):

mercurial/metadata.py

copied = ctx[dst].renamed() copied = ctx[dst].renamed()
if not copied: if not copied:
continue continue
src, srcnode = copied src, srcnode = copied
if src in p1 and p1[src].filenode() == srcnode: if src in p1 and p1[src].filenode() == srcnode:
p1copies[dst] = src p1copies[dst] = src
elif src in p2 and p2[src].filenode() == srcnode: elif src in p2 and p2[src].filenode() == srcnode:
p2copies[dst] = src p2copies[dst] = src
else:
# This case should only happen with experimental.copies.keep-old
p1copies[dst] = src
return p1copies, p2copies return p1copies, p2copies
def encodecopies(files, copies): def encodecopies(files, copies):
items = [] items = []
for i, dst in enumerate(files): for i, dst in enumerate(files):
if dst in copies: if dst in copies:
items.append(b'%d\0%s' % (i, copies[dst])) items.append(b'%d\0%s' % (i, copies[dst]))

tests/test-copies.t

Test rebasing a commit after amending rename of added file into its parent Test rebasing a commit after amending rename of added file into its parent
(issue 5457) (issue 5457)
$ cat << EOF >> $HGRCPATH $ cat << EOF >> $HGRCPATH
> [experimental] > [experimental]
> evolution.createmarkers = True > evolution.createmarkers = True
> evolution.allowunstable = True > evolution.allowunstable = True
> copies.keep-old = True
> EOF > EOF
$ newrepo $ newrepo
$ echo a > a $ echo a > a
$ hg add a $ hg add a
$ hg ci -m "added a" $ hg ci -m "added a"
$ echo b > b $ echo b > b
$ hg add b $ hg add b
$ hg ci -m "added b" $ hg ci -m "added b"
$ echo modified > b $ echo modified > b
$ hg ci -m "modified b" $ hg ci -m "modified b"
$ hg co '.^' $ hg co '.^'
1 files updated, 0 files merged, 0 files removed, 0 files unresolved 1 files updated, 0 files merged, 0 files removed, 0 files unresolved
$ hg mv b c $ hg mv b c
$ hg ci --amend -m "added c" $ hg ci --amend -m "added c"
1 new orphan changesets 1 new orphan changesets
#if no-filelog no-compatibility
$ hg rebase -s 'desc("modified b")' -d . $ hg rebase -s 'desc("modified b")' -d .
rebasing 2:e3e0011b43ad "modified b" (changeset !) rebasing 2:e3e0011b43ad "modified b" (changeset !)
rebasing 2:2612b8963c3f "modified b" (no-changeset !) rebasing 2:2612b8963c3f "modified b" (no-changeset !)
merging c and b to c
#else
$ hg rebase -s 'desc("modified b")' -d .
rebasing 2:2612b8963c3f "modified b"
file 'b' was deleted in local [dest] but was modified in other [source]. file 'b' was deleted in local [dest] but was modified in other [source].
You can use (c)hanged version, leave (d)eleted, or leave (u)nresolved. You can use (c)hanged version, leave (d)eleted, or leave (u)nresolved.
What do you want to do? u What do you want to do? u
unresolved conflicts (see 'hg resolve', then 'hg rebase --continue') unresolved conflicts (see 'hg resolve', then 'hg rebase --continue')
[240] [240]
#endif