Mercurial > hg
view tests/test-amend.t @ 46607:e9901d01d135
revlog: add a mechanism to verify expected file position before appending
If someone uses `hg debuglocks`, or some non-hg process writes to the .hg
directory without respecting the locks, or if the repo's on a networked
filesystem, it's possible for the revlog code to write out corrupted data.
The form of this corruption can vary depending on what data was written and how
that happened. We are in the "networked filesystem" case (though I've had users
also do this to themselves with the "`hg debuglocks`" scenario), and most often
see this with the changelog. What ends up happening is we produce two items
(let's call them rev1 and rev2) in the .i file that have the same linkrev,
baserev, and offset into the .d file, while the data in the .d file is appended
properly. rev2's compressed_size is accurate for rev2, but when we go to
decompress the data in the .d file, we use the offset that's recorded in the
index file, which is the same as rev1, and attempt to decompress
rev2.compressed_size bytes of rev1's data. This usually does not succeed. :)
When using inline data, this also fails, though I haven't investigated why too
closely. This shows up as a "patch decode" error. I believe what's happening
there is that we're basically ignoring the offset field, getting the data
properly, but since baserev != rev, it thinks this is a delta based on rev
(instead of a full text) and can't actually apply it as such.
For now, I'm going to make this an optional component and default it to entirely
off. I may increase the default severity of this in the future, once I've
enabled it for my users and we gain more experience with it. Luckily, most of my
users have a versioned filesystem and can roll back to before the corruption has
been written, it's just a hassle to do so and not everyone knows how (so it's a
support burden). Users on other filesystems will not have that luxury, and this
can cause them to have a corrupted repository that they are unlikely to know how
to resolve, and they'll see this as a data-loss event. Refusing to create the
corruption is a much better user experience.
This mechanism is not perfect. There may be false-negatives (racy writes that
are not detected). There should not be any false-positives (non-racy writes that
are detected as such). This is not a mechanism that makes putting a repo on a
networked filesystem "safe" or "supported", just *less* likely to cause
corruption.
Differential Revision: https://phab.mercurial-scm.org/D9952
author | Kyle Lippincott <spectral@google.com> |
---|---|
date | Wed, 03 Feb 2021 16:33:10 -0800 |
parents | b4694ef45db5 |
children | 7a90fddb13b0 |
line wrap: on
line source
#testcases obsstore-off obsstore-on $ cat << EOF >> $HGRCPATH > [extensions] > amend= > debugdrawdag=$TESTDIR/drawdag.py > [diff] > git=1 > EOF #if obsstore-on $ cat << EOF >> $HGRCPATH > [experimental] > evolution.createmarkers=True > EOF #endif Basic amend $ hg init repo1 $ cd repo1 $ hg debugdrawdag <<'EOS' > B > | > A > EOS $ hg update B -q $ echo 2 >> B $ hg amend saved backup bundle to $TESTTMP/repo1/.hg/strip-backup/112478962961-7e959a55-amend.hg (obsstore-off !) #if obsstore-off $ hg log -p -G --hidden -T '{rev} {node|short} {desc}\n' @ 1 be169c7e8dbe B | diff --git a/B b/B | new file mode 100644 | --- /dev/null | +++ b/B | @@ -0,0 +1,1 @@ | +B2 | o 0 426bada5c675 A diff --git a/A b/A new file mode 100644 --- /dev/null +++ b/A @@ -0,0 +1,1 @@ +A \ No newline at end of file #else $ hg log -p -G --hidden -T '{rev} {node|short} {desc}\n' @ 2 be169c7e8dbe B | diff --git a/B b/B | new file mode 100644 | --- /dev/null | +++ b/B | @@ -0,0 +1,1 @@ | +B2 | | x 1 112478962961 B |/ diff --git a/B b/B | new file mode 100644 | --- /dev/null | +++ b/B | @@ -0,0 +1,1 @@ | +B | \ No newline at end of file | o 0 426bada5c675 A diff --git a/A b/A new file mode 100644 --- /dev/null +++ b/A @@ -0,0 +1,1 @@ +A \ No newline at end of file #endif Nothing changed $ hg amend nothing changed [1] $ hg amend -d "0 0" nothing changed [1] $ hg amend -d "Thu Jan 01 00:00:00 1970 UTC" nothing changed [1] #if obsstore-on $ hg init repo-merge-state $ cd repo-merge-state $ echo a > f $ hg ci -Aqm a $ echo b > f $ hg ci -Aqm b $ echo c > f $ hg co -m '.^' merging f warning: conflicts while merging f! (edit, then use 'hg resolve --mark') 0 files updated, 0 files merged, 0 files removed, 1 files unresolved use 'hg resolve' to retry unresolved file merges [1] $ echo d > f $ hg resolve -m f (no more unresolved files) $ hg ci --amend --config experimental.evolution.allowunstable=True 1 new orphan changesets $ hg resolve -l $ cd .. #endif Matcher and metadata options $ echo 3 > C $ echo 4 > D $ hg add C D $ hg amend -m NEWMESSAGE -I C saved backup bundle to $TESTTMP/repo1/.hg/strip-backup/be169c7e8dbe-7684ddc5-amend.hg (obsstore-off !) $ hg log -r . -T '{node|short} {desc} {files}\n' c7ba14d9075b NEWMESSAGE B C $ echo 5 > E $ rm C $ hg amend -d '2000 1000' -u 'Foo <foo@example.com>' -A C D saved backup bundle to $TESTTMP/repo1/.hg/strip-backup/c7ba14d9075b-b3e76daa-amend.hg (obsstore-off !) $ hg log -r . -T '{node|short} {desc} {files} {author} {date}\n' 14f6c4bcc865 NEWMESSAGE B D Foo <foo@example.com> 2000.01000 Amend with editor $ cat > $TESTTMP/prefix.sh <<'EOF' > printf 'EDITED: ' > $TESTTMP/msg > cat "$1" >> $TESTTMP/msg > mv $TESTTMP/msg "$1" > EOF $ chmod +x $TESTTMP/prefix.sh $ HGEDITOR="sh $TESTTMP/prefix.sh" hg amend --edit saved backup bundle to $TESTTMP/repo1/.hg/strip-backup/14f6c4bcc865-6591f15d-amend.hg (obsstore-off !) $ hg log -r . -T '{node|short} {desc}\n' 298f085230c3 EDITED: NEWMESSAGE $ HGEDITOR="sh $TESTTMP/prefix.sh" hg amend -e -m MSG saved backup bundle to $TESTTMP/repo1/.hg/strip-backup/298f085230c3-d81a6ad3-amend.hg (obsstore-off !) $ hg log -r . -T '{node|short} {desc}\n' 974f07f28537 EDITED: MSG $ echo FOO > $TESTTMP/msg $ hg amend -l $TESTTMP/msg -m BAR abort: cannot specify both --message and --logfile [10] $ hg amend -l $TESTTMP/msg saved backup bundle to $TESTTMP/repo1/.hg/strip-backup/974f07f28537-edb6470a-amend.hg (obsstore-off !) $ hg log -r . -T '{node|short} {desc}\n' 507be9bdac71 FOO Interactive mode $ touch F G $ hg add F G $ cat <<EOS | hg amend -i --config ui.interactive=1 > y > n > EOS diff --git a/F b/F new file mode 100644 examine changes to 'F'? (enter ? for help) [Ynesfdaq?] y diff --git a/G b/G new file mode 100644 examine changes to 'G'? (enter ? for help) [Ynesfdaq?] n saved backup bundle to $TESTTMP/repo1/.hg/strip-backup/507be9bdac71-c8077452-amend.hg (obsstore-off !) $ hg log -r . -T '{files}\n' B D F Amend in the middle of a stack $ hg init $TESTTMP/repo2 $ cd $TESTTMP/repo2 $ hg debugdrawdag <<'EOS' > C > | > B > | > A > EOS $ hg update -q B $ echo 2 >> B $ hg amend abort: cannot amend changeset with children [10] #if obsstore-on With allowunstable, amend could work in the middle of a stack $ cat >> $HGRCPATH <<EOF > [experimental] > evolution.createmarkers=True > evolution.allowunstable=True > EOF $ hg amend 1 new orphan changesets $ hg log -T '{rev} {node|short} {desc}\n' -G @ 3 be169c7e8dbe B | | * 2 26805aba1e60 C | | | x 1 112478962961 B |/ o 0 426bada5c675 A Checking the note stored in the obsmarker $ echo foo > bar $ hg add bar $ hg amend --note 'yyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyy' abort: cannot store a note of more than 255 bytes [10] $ hg amend --note "adding bar" $ hg debugobsolete -r . 112478962961147124edd43549aedd1a335e44bf be169c7e8dbe21cd10b3d79691cbe7f241e3c21c 0 (Thu Jan 01 00:00:00 1970 +0000) {'ef1': '8', 'operation': 'amend', 'user': 'test'} be169c7e8dbe21cd10b3d79691cbe7f241e3c21c 16084da537dd8f84cfdb3055c633772269d62e1b 0 (Thu Jan 01 00:00:00 1970 +0000) {'ef1': '8', 'note': 'adding bar', 'operation': 'amend', 'user': 'test'} #endif Cannot amend public changeset $ hg phase -r A --public $ hg update -C -q A $ hg amend -m AMEND abort: cannot amend public changesets (see 'hg help phases' for details) [10] Amend a merge changeset $ hg init $TESTTMP/repo3 $ cd $TESTTMP/repo3 $ hg debugdrawdag <<'EOS' > C > /| > A B > EOS $ hg update -q C $ hg amend -m FOO saved backup bundle to $TESTTMP/repo3/.hg/strip-backup/a35c07e8a2a4-15ff4612-amend.hg (obsstore-off !) $ rm .hg/localtags $ hg log -G -T '{desc}\n' @ FOO |\ | o B | o A More complete test for status changes (issue5732) ------------------------------------------------- Generates history of files having 3 states, r0_r1_wc: r0: ground (content/missing) r1: old state to be amended (content/missing, where missing means removed) wc: changes to be included in r1 (content/missing-tracked/untracked) $ hg init $TESTTMP/wcstates $ cd $TESTTMP/wcstates $ "$PYTHON" $TESTDIR/generate-working-copy-states.py state 2 1 $ hg addremove -q --similarity 0 $ hg commit -m0 $ "$PYTHON" $TESTDIR/generate-working-copy-states.py state 2 2 $ hg addremove -q --similarity 0 $ hg commit -m1 $ "$PYTHON" $TESTDIR/generate-working-copy-states.py state 2 wc $ hg addremove -q --similarity 0 $ hg forget *_*_*-untracked $ rm *_*_missing-* amend r1 to include wc changes $ hg amend saved backup bundle to * (glob) (obsstore-off !) clean/modified/removed/added states of the amended revision $ hg status --all --change . 'glob:content1_*_content1-tracked' C content1_content1_content1-tracked C content1_content2_content1-tracked C content1_missing_content1-tracked $ hg status --all --change . 'glob:content1_*_content[23]-tracked' M content1_content1_content3-tracked M content1_content2_content2-tracked M content1_content2_content3-tracked M content1_missing_content3-tracked $ hg status --all --change . 'glob:content1_*_missing-tracked' M content1_content2_missing-tracked R content1_missing_missing-tracked C content1_content1_missing-tracked $ hg status --all --change . 'glob:content1_*_*-untracked' R content1_content1_content1-untracked R content1_content1_content3-untracked R content1_content1_missing-untracked R content1_content2_content1-untracked R content1_content2_content2-untracked R content1_content2_content3-untracked R content1_content2_missing-untracked R content1_missing_content1-untracked R content1_missing_content3-untracked R content1_missing_missing-untracked $ hg status --all --change . 'glob:missing_content2_*' A missing_content2_content2-tracked A missing_content2_content3-tracked A missing_content2_missing-tracked $ hg status --all --change . 'glob:missing_missing_*' A missing_missing_content3-tracked working directory should be all clean (with some missing/untracked files) $ hg status --all 'glob:*_content?-tracked' C content1_content1_content1-tracked C content1_content1_content3-tracked C content1_content2_content1-tracked C content1_content2_content2-tracked C content1_content2_content3-tracked C content1_missing_content1-tracked C content1_missing_content3-tracked C missing_content2_content2-tracked C missing_content2_content3-tracked C missing_missing_content3-tracked $ hg status --all 'glob:*_missing-tracked' ! content1_content1_missing-tracked ! content1_content2_missing-tracked ! content1_missing_missing-tracked ! missing_content2_missing-tracked ! missing_missing_missing-tracked $ hg status --all 'glob:*-untracked' ? content1_content1_content1-untracked ? content1_content1_content3-untracked ? content1_content2_content1-untracked ? content1_content2_content2-untracked ? content1_content2_content3-untracked ? content1_missing_content1-untracked ? content1_missing_content3-untracked ? missing_content2_content2-untracked ? missing_content2_content3-untracked ? missing_missing_content3-untracked ================================= Test backup-bundle config option| ================================= $ hg init $TESTTMP/repo4 $ cd $TESTTMP/repo4 $ echo a>a $ hg ci -Aqma $ echo oops>b $ hg ci -Aqm "b" $ echo partiallyfixed > b #if obsstore-off $ hg amend saved backup bundle to $TESTTMP/repo4/.hg/strip-backup/95e899acf2ce-f11cb050-amend.hg When backup-bundle config option is set: $ cat << EOF >> $HGRCPATH > [rewrite] > backup-bundle = False > EOF $ echo fixed > b $ hg amend #else $ hg amend When backup-bundle config option is set: $ cat << EOF >> $HGRCPATH > [rewrite] > backup-bundle = False > EOF $ echo fixed > b $ hg amend #endif ========================================== Test update-timestamp config option| ========================================== $ cat >> $HGRCPATH << EOF > [extensions] > amend= > mockmakedate = $TESTDIR/mockmakedate.py > EOF $ hg init $TESTTMP/repo5 $ cd $TESTTMP/repo5 $ cat <<'EOF' >> .hg/hgrc > [command-templates] > log = 'user: {user} > date: {date|date} > summary: {desc|firstline}\n' > EOF $ echo a>a $ hg ci -Am 'commit 1' adding a When updatetimestamp is False $ hg amend --date '1997-1-1 0:1' $ hg log --limit 1 user: test date: Wed Jan 01 00:01:00 1997 +0000 summary: commit 1 When update-timestamp is True and no other change than the date $ hg amend --config rewrite.update-timestamp=True nothing changed [1] $ hg log --limit 1 user: test date: Wed Jan 01 00:01:00 1997 +0000 summary: commit 1 When update-timestamp is True and there is other change than the date $ hg amend --user foobar --config rewrite.update-timestamp=True $ hg log --limit 1 user: foobar date: Thu Jan 01 00:00:02 1970 +0000 summary: commit 1 When date option is applicable and update-timestamp is True $ hg amend --date '1998-1-1 0:1' --config rewrite.update-timestamp=True $ hg log --limit 1 user: foobar date: Thu Jan 01 00:01:00 1998 +0000 summary: commit 1 Unlike rewrite.update-timestamp, -D/--currentdate always updates the timestamp $ hg amend -D $ hg log --limit 1 user: foobar date: Thu Jan 01 00:00:04 1970 +0000 summary: commit 1 $ hg amend -D --config rewrite.update-timestamp=True $ hg log --limit 1 user: foobar date: Thu Jan 01 00:00:05 1970 +0000 summary: commit 1 rewrite.update-timestamp can be negated by --no-currentdate $ hg amend --config rewrite.update-timestamp=True --no-currentdate -u baz $ hg log --limit 1 user: baz date: Thu Jan 01 00:00:05 1970 +0000 summary: commit 1 Bad combination of date options: $ hg amend -D --date '0 0' abort: cannot specify both --date and --currentdate [10] Close branch $ hg amend --secret --close-branch $ hg log --limit 1 -T 'close={get(extras, "close")}\nphase={phase}\n' close=1 phase=secret $ cd .. Corner case of amend from issue6157: - working copy parent has a change to file `a` - working copy has the inverse change - we amend the working copy parent for files other than `a` hg used to include the changes to `a` anyway. $ hg init 6157; cd 6157 $ echo a > a; echo b > b; hg commit -qAm_ $ echo a2 > a; hg commit -qm_ $ hg diff --stat -c . a | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) $ echo a > a; echo b2 > b; hg amend -q b $ hg diff --stat -c . a | 2 +- b | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) Modifying a file while the editor is open can cause dirstate corruption (issue6233) $ cd $TESTTMP $ hg init modify-during-amend; cd modify-during-amend $ echo r0 > foo; hg commit -qAm "r0" $ echo alpha > foo; hg commit -qm "alpha" $ echo beta >> foo $ cat > $TESTTMP/touchy_editor.sh <<EOF > sleep 1 > echo delta >> "$TESTTMP/modify-during-amend/foo" > sleep 1 > echo hi > "\$1" > sleep 1 > EOF $ HGEDITOR="sh $TESTTMP/touchy_editor.sh" hg commit --amend $ if (hg diff -c . | grep 'delta' >/dev/null) || [ -n "$(hg status)" ]; then > echo "OK." > else > echo "Bug detected. 'delta' is not part of the commit OR the wdir" > echo "Diff and status before rebuild:" > hg diff > hg status > hg debugrebuilddirstate > echo "Diff and status after rebuild:" > hg diff > hg status > fi OK.