Mercurial > hg
view tests/test-share.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 | 4b0192f592cf |
children | 8be95673eb8a |
line wrap: on
line source
#testcases safe normal #if safe $ echo "[format]" >> $HGRCPATH $ echo "use-share-safe = True" >> $HGRCPATH #endif $ echo "[extensions]" >> $HGRCPATH $ echo "share = " >> $HGRCPATH prepare repo1 $ hg init repo1 $ cd repo1 $ echo a > a $ hg commit -A -m'init' adding a share it $ cd .. $ hg share repo1 repo2 updating working directory 1 files updated, 0 files merged, 0 files removed, 0 files unresolved share shouldn't have a store dir $ cd repo2 $ test -d .hg/store [1] $ hg root -Tjson | sed 's|\\\\|\\|g' [ { "hgpath": "$TESTTMP/repo2/.hg", "reporoot": "$TESTTMP/repo2", "storepath": "$TESTTMP/repo1/.hg/store" } ] share shouldn't have a full cache dir, original repo should $ hg branches default 0:d3873e73d99e $ hg tags tip 0:d3873e73d99e $ test -d .hg/cache [1] $ ls -1 .hg/wcache || true checkisexec (execbit !) checklink (symlink !) checklink-target (symlink !) manifestfulltextcache (reporevlogstore !) $ ls -1 ../repo1/.hg/cache branch2-served rbc-names-v1 rbc-revs-v1 tags2-visible Cloning a shared repo should pick up the full cache dir on the other hand. $ hg clone . ../repo2-clone updating to branch default 1 files updated, 0 files merged, 0 files removed, 0 files unresolved $ ls -1 ../repo2-clone/.hg/cache branch2-base branch2-immutable branch2-served branch2-served.hidden branch2-visible branch2-visible-hidden hgtagsfnodes1 rbc-names-v1 rbc-revs-v1 tags2 tags2-served tags2-visible Some sed versions appends newline, some don't, and some just fails $ cat .hg/sharedpath; echo $TESTTMP/repo1/.hg trailing newline on .hg/sharedpath is ok $ hg tip -q 0:d3873e73d99e $ echo '' >> .hg/sharedpath $ cat .hg/sharedpath $TESTTMP/repo1/.hg $ hg tip -q 0:d3873e73d99e commit in shared clone $ echo a >> a $ hg commit -m'change in shared clone' check original $ cd ../repo1 $ hg log changeset: 1:8af4dc49db9e tag: tip user: test date: Thu Jan 01 00:00:00 1970 +0000 summary: change in shared clone changeset: 0:d3873e73d99e user: test date: Thu Jan 01 00:00:00 1970 +0000 summary: init $ hg update 1 files updated, 0 files merged, 0 files removed, 0 files unresolved $ cat a # should be two lines of "a" a a commit in original $ echo b > b $ hg commit -A -m'another file' adding b check in shared clone $ cd ../repo2 $ hg log changeset: 2:c2e0ac586386 tag: tip user: test date: Thu Jan 01 00:00:00 1970 +0000 summary: another file changeset: 1:8af4dc49db9e user: test date: Thu Jan 01 00:00:00 1970 +0000 summary: change in shared clone changeset: 0:d3873e73d99e user: test date: Thu Jan 01 00:00:00 1970 +0000 summary: init $ hg update 1 files updated, 0 files merged, 0 files removed, 0 files unresolved $ cat b # should exist with one "b" b hg serve shared clone $ hg serve -n test -p $HGPORT -d --pid-file=hg.pid $ cat hg.pid >> $DAEMON_PIDS $ get-with-headers.py localhost:$HGPORT 'raw-file/' 200 Script output follows -rw-r--r-- 4 a -rw-r--r-- 2 b Cloning a shared repo via bundle2 results in a non-shared clone $ cd .. $ hg clone -q --stream --config ui.ssh="\"$PYTHON\" \"$TESTDIR/dummyssh\"" ssh://user@dummy/`pwd`/repo2 cloned-via-bundle2 $ cat ./cloned-via-bundle2/.hg/requires | grep "shared" [1] $ hg id --cwd cloned-via-bundle2 -r tip c2e0ac586386 tip $ cd repo2 test unshare command $ hg unshare $ test -d .hg/store $ test -f .hg/sharedpath [1] $ grep shared .hg/requires [1] $ hg unshare abort: this is not a shared repo [255] check that a change does not propagate $ echo b >> b $ hg commit -m'change in unshared' $ cd ../repo1 $ hg id -r tip c2e0ac586386 tip $ cd .. non largefiles repos won't enable largefiles $ hg share --config extensions.largefiles= repo2 sharedrepo The fsmonitor extension is incompatible with the largefiles extension and has been disabled. (fsmonitor !) The fsmonitor extension is incompatible with the largefiles extension and has been disabled. (fsmonitor !) updating working directory 2 files updated, 0 files merged, 0 files removed, 0 files unresolved $ [ -f sharedrepo/.hg/hgrc ] [1] test shared clones using relative paths work $ mkdir thisdir $ hg init thisdir/orig $ hg share -U thisdir/orig thisdir/abs $ hg share -U --relative thisdir/abs thisdir/rel $ cat thisdir/rel/.hg/sharedpath ../../orig/.hg (no-eol) $ grep shared thisdir/*/.hg/requires thisdir/abs/.hg/requires:shared thisdir/rel/.hg/requires:relshared thisdir/rel/.hg/requires:shared test that relative shared paths aren't relative to $PWD $ cd thisdir $ hg -R rel root $TESTTMP/thisdir/rel $ cd .. now test that relative paths really are relative, survive across renames and changes of PWD $ hg -R thisdir/abs root $TESTTMP/thisdir/abs $ hg -R thisdir/rel root $TESTTMP/thisdir/rel $ mv thisdir thatdir $ hg -R thatdir/abs root abort: .hg/sharedpath points to nonexistent directory $TESTTMP/thisdir/orig/.hg [255] $ hg -R thatdir/rel root $TESTTMP/thatdir/rel test unshare relshared repo $ cd thatdir/rel $ hg unshare $ test -d .hg/store $ test -f .hg/sharedpath [1] $ grep shared .hg/requires [1] $ hg unshare abort: this is not a shared repo [255] $ cd ../.. $ rm -r thatdir Demonstrate buggy behavior around requirements validation See comment in localrepo.py:makelocalrepository() for more. $ hg init sharenewrequires $ hg share sharenewrequires shareoldrequires updating working directory 0 files updated, 0 files merged, 0 files removed, 0 files unresolved $ cat >> sharenewrequires/.hg/requires << EOF > missing-requirement > EOF We cannot open the repo with the unknown requirement $ hg -R sharenewrequires status abort: repository requires features unknown to this Mercurial: missing-requirement (see https://mercurial-scm.org/wiki/MissingRequirement for more information) [255] BUG: we don't get the same error when opening the shared repo pointing to it $ hg -R shareoldrequires status Explicitly kill daemons to let the test exit on Windows $ killdaemons.py Test sharing a repository which was created with store requirement disable $ hg init nostore --config format.usestore=false ignoring enabled 'format.use-share-safe' config because it is incompatible with disabled 'format.usestore' config (safe !) $ hg share nostore sharednostore abort: cannot create shared repository as source was created with 'format.usestore' config disabled [255]