|
Ever since the split index feature was introduced [1], refreshing a
split index is prone to a variant of the classic racy git problem.
Consider the following sequence of commands updating the split index
when the shared index contains a racily clean cache entry, i.e. an
entry whose cached stat data matches with the corresponding file in
the worktree and the cached mtime matches that of the index:
echo "cached content" >file
git update-index --split-index --add file
echo "dirty worktree" >file # size stays the same!
# ... wait ...
git update-index --add other-file
Normally, when a non-split index is updated, then do_write_index()
(the function responsible for writing all kinds of indexes, "regular",
split, and shared) recognizes racily clean cache entries, and writes
them with smudged stat data, i.e. with file size set to 0. When
subsequent git commands read the index, they will notice that the
smudged stat data doesn't match with the file in the worktree, and
then go on to check the file's content and notice its dirtiness.
In the above example, however, in the second 'git update-index'
prepare_to_write_split_index() decides which cache entries stored only
in the shared index should be replaced in the new split index. Alas,
this function never looks out for racily clean cache entries, and
since the file's stat data in the worktree hasn't changed since the
shared index was written, it won't be replaced in the new split index.
Consequently, do_write_index() doesn't even get this racily clean
cache entry, and can't smudge its stat data. Subsequent git commands
will then see that the index has more recent mtime than the file and
that the (not smudged) cached stat data still matches with the file in
the worktree, and, ultimately, will erroneously consider the file
clean.
Modify prepare_to_write_split_index() to recognize racily clean cache
entries, and mark them to be added to the split index. Note that
there are two places where it should check raciness: first those cache
entries that are only stored in the shared index, and then those that
have been copied by unpack_trees() from the shared index while it
constructed a new index. This way do_write_index() will get these
racily clean cache entries as well, and will then write them with
smudged stat data to the new split index.
This change makes all tests in 't1701-racy-split-index.sh' pass, so
flip the two 'test_expect_failure' tests to success. Also add the '#'
(as in nr. of trial) to those tests' description that were omitted
when the tests expected failure.
Note that after this change if the index is split when it contains a
racily clean cache entry, then a smudged cache entry will be written
both to the new shared and to the new split indexes. This doesn't
affect regular git commands: as far as they are concerned this is just
an entry in the split index replacing an outdated entry in the shared
index. It did affect a few tests in 't1700-split-index.sh', though,
because they actually check which entries are stored in the split
index; a previous patch in this series has already made the necessary
adjustments in 't1700'. And racily clean cache entries and index
splitting are rare enough to not worry about the resulting duplicated
smudged cache entries, and the additional complexity required to
prevent them is not worth it.
Several tests failed occasionally when the test suite was run with
'GIT_TEST_SPLIT_INDEX=yes'. Here are those that I managed to trace
back to this racy split index problem, starting with those failing
more frequently, with a link to a failing Travis CI build job for
each. The highlighted line [2] shows when the racy file was written,
which is not always in the failing test but in a preceeding setup
test.
t3903-stash.sh:
https://travis-ci.org/git/git/jobs/385542084#L5858
t4024-diff-optimize-common.sh:
https://travis-ci.org/git/git/jobs/386531969#L3174
t4015-diff-whitespace.sh:
https://travis-ci.org/git/git/jobs/360797600#L8215
t2200-add-update.sh:
https://travis-ci.org/git/git/jobs/382543426#L3051
t0090-cache-tree.sh:
https://travis-ci.org/git/git/jobs/416583010#L3679
There might be others, e.g. perhaps 't1000-read-tree-m-3way.sh' and
others using 'lib-read-tree-m-3way.sh', but I couldn't confirm yet.
[1] In the branch leading to the merge commit v2.1.0-rc0~45 (Merge
branch 'nd/split-index', 2014-07-16).
[2] Note that those highlighted lines are in the 'after failure' fold,
and your browser might unhelpfully fold it up before you could
take a good look.
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
|
Ever since the split index feature was introduced [1], refreshing a
split index is prone to a variant of the classic racy git problem.
There are a couple of unrelated tests in the test suite that
occasionally fail when run with 'GIT_TEST_SPLIT_INDEX=yes', but
't1700-split-index.sh', the only test script focusing solely on split
index, has never noticed this issue, because it only cares about how
the index is split under various circumstances and all the different
ways to turn the split index feature on and off.
Add a dedicated test script 't1701-racy-split-index.sh' to exercise
the split index feature in racy situations as well; kind of a
"t0010-racy-git.sh for split index" but with modern style (the tests
do everything in &&-chained list of commands in 'test_expect_...'
blocks, and use 'test_cmp' for more informative output on failure).
The tests cover the following sequences of index splitting, updating,
and racy file modifications, with the last two cases demonstrating the
racy split index problem:
1. Split the index while adding a racily clean file:
echo "cached content" >file
git update-index --split-index --add file
echo "dirty worktree" >file # size stays the same
This case already works properly. Even though the cache entry's
stat data matches with the modifid file in the worktree,
subsequent git commands will notice that the (split) index and
the file have the same mtime, and then will go on to check the
file's content and notice its dirtiness.
2. Add a racily clean file to an already split index:
git update-index --split-index
echo "cached content" >file
git update-index --add file
echo "dirty worktree" >file
This case already works properly. After the second 'git
update-index' writes the newly added file's cache entry to the
new split index, it basically works in the same way as case #1.
3. Split the index when it (i.e. the not yet splitted index)
contains a racily clean cache entry, i.e. an entry whose cached
stat data matches with the corresponding file in the worktree and
the cached mtime matches that of the index:
echo "cached content" >file
git update-index --add file
echo "dirty worktree" >file
# ... wait ...
git update-index --split-index --add other-file
This case already works properly. The shared index is written by
do_write_index(), i.e. the same function that is responsible for
writing "regular" and split indexes as well. This function
cleverly notices the racily clean cache entry, and writes the
entry to the new shared index with smudged stat data, i.e. file
size set to 0. When subsequent git commands read the index, they
will notice that the smudged stat data doesn't match with the
file in the worktree, and then go on to check the file's content
and notice its dirtiness.
4. Update the split index when it contains a racily clean cache
entry:
git update-index --split-index
echo "cached content" >file
git update-index --add file
echo "dirty worktree" >file
# ... wait ...
git update-index --add other-file
This case already works properly. After the second 'git
update-index' the newly added file's cache entry is only stored
in the split index. If a cache entry is present in the split
index (even if it is a replacement of an outdated entry in the
shared index), then it will always be included in the new split
index on subsequent split index updates (until the file is
removed or a new shared index is written), independently from
whether the entry is racily clean or not. When do_write_index()
writes the new split index, it notices the racily clean cache
entry, and smudges its stat date. Subsequent git commands
reading the index will notice the smudged stat data and then go
on to check the file's content and notice its dirtiness.
5. Update the split index when a racily clean cache entry is stored
only in the shared index:
echo "cached content" >file
git update-index --split-index --add file
echo "dirty worktree" >file
# ... wait ...
git update-index --add other-file
This case fails due to the racy split index problem. In the
second 'git update-index' prepare_to_write_split_index() decides,
among other things, which cache entries stored only in the shared
index should be replaced in the new split index. Alas, this
function never looks out for racily clean cache entries, and
since the file's stat data in the worktree hasn't changed since
the shared index was written, the entry won't be replaced in the
new split index. Consequently, do_write_index() doesn't even get
this racily clean cache entry, and can't smudge its stat data.
Subsequent git commands will then see that the index has more
recent mtime than the file and that the (not smudged) cached stat
data still matches with the file in the worktree, and,
ultimately, will erroneously consider the file clean.
6. Update the split index after unpack_trees() copied a racily clean
cache entry from the shared index:
echo "cached content" >file
git update-index --split-index --add file
echo "dirty worktree" >file
# ... wait ...
git read-tree -m HEAD
This case fails due to the racy split index problem. This
basically fails for the same reason as case #5 above, but there
is one important difference, which warrants the dedicated test.
While that second 'git update-index' in case #5 updates
index_state in place, in this case 'git read-tree -m' calls
unpack_trees(), which throws out the entire index, and constructs
a new one from the (potentially updated) copies of the original's
cache entries. Consequently, when prepare_to_write_split_index()
gets to work on this reconstructed index, it takes a different
code path than in case #5 when deciding which cache entries in
the shared index should be replaced. The result is the same,
though: the racily clean cache entry goes unnoticed, it isn't
added to the split index with smudged stat data, and subsequent
git commands will then erroneously consider the file clean.
Note that in the last two 'test_expect_failure' cases I omitted the
'#' (as in nr. of trial) from the tests' description on purpose for
now, as it breakes the TAP output [2]; it will be added at the end of
the series, when those two tests will be flipped to
'test_expect_success'.
[1] In the branch leading to the merge commit v2.1.0-rc0~45 (Merge
branch 'nd/split-index', 2014-07-16).
[2] In the TAP output a '#' should separate the test's description
from the TODO directive emitted by 'test_expect_failure'. The
additional '#' in "#$trial" interferes with this, the test harness
won't recognize the TODO directive, and will report that those
tests failed unexpectedly.
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
SZEDER Gábor