Change all in-tree users of the string_list_init(LIST, BOOL) API to
use string_list_init_{nodup,dup}(LIST) instead.
As noted in the preceding commit let's leave the now-unused
string_list_init() wrapper in-place for any in-flight users, it can be
removed at some later date.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Use -1 as error return value throughout.
This removes spurious differences in the GIT_TRACE_REFS output, depending on the
ref storage backend active.
Before, the cached ref_iterator (but only that iterator!) would return
peel_object() output directly. No callers relied on the peel_status values
beyond success/failure. All calls to these functions go through
peel_iterated_oid(), which returns peel_object() as a fallback, but also
squashing the error values.
The iteration interface already passes REF_ISSYMREF and REF_ISBROKEN through the
flags argument, so the additional error values in enum peel_status provide no
value.
The ref iteration interface provides a separate peel() function because certain
formats (eg. packed-refs and reftable) can store the peeled object next to the
tag SHA1. Passing the peeled SHA1 as an optional argument to each_ref_fn maps
more naturally to the implementation of ref databases. Changing the code in this
way is left for a future refactoring.
Signed-off-by: Han-Wen Nienhuys <hanwen@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add and apply a semantic patch for converting code that open-codes
CALLOC_ARRAY to use it instead. It shortens the code and infers the
element size automatically.
Signed-off-by: René Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When a packed ref is deleted, the whole packed-refs file is
rewritten to omit the ref that no longer exists. However if another
gc command is running and calls `pack-refs --all` simultaneously,
there is a chance that a ref that was just updated lose the newly
created commits.
Through these steps, losing commits on newly updated refs can be
demonstrated:
# step 1: compile git without `USE_NSEC` option
Some kernel releases do enable it by default while some do
not. And if we compile git without `USE_NSEC`, it will be easier
demonstrated by the following steps.
# step 2: setup a repository and add the first commit
git init repo &&
(cd repo &&
git config core.logallrefupdates true &&
git commit --allow-empty -m foo)
# step 3: in one terminal, repack the refs repeatedly
cd repo &&
while true
do
git pack-refs --all
done
# step 4: in another terminal, simultaneously update the
# master with update-ref, and create and delete an
# unrelated ref also with update-ref
cd repo &&
while true
do
us=$(git commit-tree -m foo -p HEAD HEAD^{tree}) &&
git update-ref refs/heads/newbranch $us &&
git update-ref refs/heads/master $us &&
git update-ref -d refs/heads/newbranch &&
them=$(git rev-parse master) &&
if test "$them" != "$us"
then
echo >&2 "lost commit: $us"
exit 1
fi
# eye candy
printf .
done
Though we have the packed-refs lock file and loose refs lock
files to avoid updating conflicts, a ref will lost its newly
commits if racy stat-validity of `packed-refs` file happens
(which is quite same as the racy-git described in
`Documentation/technical/racy-git.txt`), the following
specific set of operations demonstrates the problem:
1. Call `pack-refs --all` to pack all the loose refs to
packed-refs, and let say the modify time of the
packed-refs is DATE_M.
2. Call `update-ref` to update a new commit to master while
it is already packed. the old value (let us call it
OID_A) remains in the packed-refs file and write the new
value (let us call it OID_B) to $GIT_DIR/refs/heads/master.
3. Call `update-ref -d` within the same DATE_M from the 1th
step to delete a different ref newbranch which is packed
in the packed-refs file. It check newbranch's oid from
packed-refs file without locking it.
Meanwhile it keeps a snapshot of the packed-refs file in
memory and record the file's attributes with the snapshot.
The oid of master in the packed-refs's snapshot is OID_A.
4. Call a new `pack-refs --all` to pack the loose refs, the
oid of master in packe-refs file is OID_B, and the loose
refs $GIT_DIR/refs/heads/master is removed. Let's say
the `pack-refs --all` is very quickly done and the new
packed-refs file's modify time is still DATE_M, and it
has the same file size, even the same inode.
5. 3th step now goes on after checking the newbranch, it
begin to rewrite the packed-refs file. After get the
lock file of packed-ref file, it checks it's on-disk
file attributes with the snapshot, suck as the timestamp,
the file size and the inode value. If they are both the
same values, and the snapshot is not refreshed.
Because the loose ref of master is removed by 4th step,
`update-ref -d` will updates the new packed-ref to disk
which contains master with the oid OID_A. So now the
newly commit OID_B of master is lost.
The best path forward is just always refreshing after take
the lock file of `packed-refs` file. Traditionally we avoided
that because refreshing it implied parsing the whole file.
But these days we mmap it, so it really is just an extra
open()/mmap() and a quick read of the header. That doesn't seem
like an outrageous cost to pay when we're already taking the lock.
Signed-off-by: Sun Chao <sunchao9@huawei.com>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Sun Chao <sunchao9@huawei.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Switch uses of GIT_SHA1_HEXSZ to use the_hash_algo so that they are
appropriate for the any given hash length.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This is the flip side of the previous two patches: checking
for a non-zero oidcmp() can be more strictly expressed as
inequality. Like those patches, we write "!= 0" in the
coccinelle transformation, which covers by isomorphism the
more common:
if (oidcmp(E1, E2))
As with the previous two patches, this patch can be achieved
almost entirely by running "make coccicheck"; the only
differences are manual line-wrap fixes to match the original
code.
There is one thing to note for anybody replicating this,
though: coccinelle 1.0.4 seems to miss the case in
builtin/tag.c, even though it's basically the same as all
the others. Running with 1.0.7 does catch this, so
presumably it's just a coccinelle bug that was fixed in the
interim.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In d8193743e0 (usage.c: add BUG() function, 2017-05-12), a new macro
was introduced to use for reporting bugs instead of die(). It was then
subsequently used to convert one single caller in 588a538ae5
(setup_git_env: convert die("BUG") to BUG(), 2017-05-12).
The cover letter of the patch series containing this patch
(cf 20170513032414.mfrwabt4hovujde2@sigill.intra.peff.net) is not
terribly clear why only one call site was converted, or what the plan
is for other, similar calls to die() to report bugs.
Let's just convert all remaining ones in one fell swoop.
This trick was performed by this invocation:
sed -i 's/die("BUG: /BUG("/g' $(git grep -l 'die("BUG' \*.c)
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit f57f37e2e1 (files-backend: remove the use of
git_path(), 2017-03-26) introduced a regression when a
relative $GIT_DIR is used in a working tree:
- when we initialize the ref backend, we make a copy of
get_git_dir(), which may be relative
- later, we may call setup_work_tree() and chdir to the
root of the working tree
- further calls to the ref code will use the stored git
directory, but relative paths will now point to the
wrong place
The new test in t1501 demonstrates one such instance (the
bug causes us to write the ref update to the nonsense
"relative/relative/.git").
Since setup_work_tree() now uses chdir_notify, we can just
ask it update our relative paths when necessary.
Reported-by: Rafael Ascensao <rafa.almas@gmail.com>
Helped-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Take a hint from commit ea68b0ce9f (hash-object: don't use mmap() for
small files, 2010-02-21) and use read() instead of mmap() for small
packed-refs files.
Signed-off-by: Kim Gybels <kgybels@infogroep.be>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We don't actually create zero-length `packed-refs` files, but they are
valid and we should handle them correctly. The old code `xmmap()`ed
such files, which led to an error when `munmap()` was called. So, if
the `packed-refs` file is empty, leave the snapshot at its zero values
and return 0 without trying to read or mmap the file.
Returning 0 also makes `create_snapshot()` exit early, which avoids
the technically undefined comparison `NULL < NULL`.
Reported-by: Kim Gybels <kgybels@infogroep.be>
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We can return an empty iterator not only if the `packed-refs` file is
missing, but also if it is empty or if there are no references whose
names succeed `prefix`. Optimize away those cases as well by moving
the call to `find_reference_location()` higher in the function and
checking whether the determined start position is the same as
`snapshot->eof`. (This is possible now because the previous commit
made `find_reference_location()` robust against empty snapshots.)
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This function had two problems if called for an empty snapshot (i.e.,
`snapshot->start == snapshot->eof == NULL`):
* It checked `NULL < NULL`, which is undefined by C (albeit highly
unlikely to fail in the real world).
* (Assuming the above comparison behaved as expected), it returned
NULL when `mustexist` was false, contrary to its docstring.
Change the check and fix the docstring.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Store a pointer to the start of the actual references within the
`packed-refs` contents rather than storing the length of the header.
This is more convenient for most users of this field.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Change `write_packed_entry()` to take `struct object_id *` rather than
`unsigned char *` arguments.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Even when we are deleting references, we needn't overwrite the
`packed-refs` file if the references that we are deleting only exist
as loose references. Implement this optimization as follows:
* Add a function `is_packed_transaction_needed()`, which checks
whether a given packed-refs transaction actually needs to be carried
out (i.e., it returns false if the transaction obviously wouldn't
have any effect). This function must be called while holding the
`packed-refs` lock to avoid races.
* Change `files_transaction_prepare()` to check whether the
packed-refs transaction is actually needed. If not, squelch it, but
continue holding the `packed-refs` lock until the end of the
transaction to avoid races.
This fixes a mild regression caused by dc39e09942 (files_ref_store:
use a transaction to update packed refs, 2017-09-08). Before that
commit, unnecessary rewrites of `packed-refs` were suppressed by
`repack_without_refs()`. But the transaction-based writing introduced
by that commit didn't perform that optimization.
Note that the pre-dc39e09942 code still had to *read* the whole
`packed-refs` file to determine that the rewrite could be skipped, so
the performance for the cases that the write could be elided was
`O(N)` in the number of packed references both before and after
dc39e09942. But after that commit the constant factor increased.
This commit reimplements the optimization of eliding unnecessary
`packed-refs` rewrites. That, plus the fact that since
cfa2e29c34 (packed_ref_store: get rid of the `ref_cache` entirely,
2017-03-17) we don't necessarily have to read the whole `packed-refs`
file at all, means that deletes of one or a few loose references can
now be done with `O(n lg N)` effort, where `n` is the number of loose
references being deleted and `N` is the total number of packed
references.
This commit fixes two tests in t1409.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Convert the unsigned char * parameter to struct object_id * for
files_read_raw_ref and packed_read_raw_ref. Update the documentation.
Switch from using get_sha1_hex and a hard-coded 40 to using
parse_oid_hex.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
reflog_expire already used struct object_id internally, but it did not
take it as a parameter. Adjust the parameter (and the callers) to pass
a pointer to struct object_id instead of a pointer to unsigned char.
Remove the temporary inserted earlier as it is no longer required.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Use strbuf_addstr() instead of strbuf_addf() for adding strings. That's
simpler and makes the intent clearer.
Patch generated by Coccinelle and contrib/coccinelle/strbuf.cocci;
adjusted indentation in refs/packed-backend.c manually.
Signed-off-by: Rene Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We've made huge changes to this file, and some of the old names and
comments are no longer very fitting. So rename a bunch of things:
* `struct packed_ref_cache` → `struct snapshot`
* `acquire_packed_ref_cache()` → `acquire_snapshot()`
* `release_packed_ref_buffer()` → `clear_snapshot_buffer()`
* `release_packed_ref_cache()` → `release_snapshot()`
* `clear_packed_ref_cache()` → `clear_snapshot()`
* `struct packed_ref_entry` → `struct snapshot_record`
* `cmp_packed_ref_entries()` → `cmp_packed_ref_records()`
* `cmp_entry_to_refname()` → `cmp_record_to_refname()`
* `sort_packed_refs()` → `sort_snapshot()`
* `read_packed_refs()` → `create_snapshot()`
* `validate_packed_ref_cache()` → `validate_snapshot()`
* `get_packed_ref_cache()` → `get_snapshot()`
* Renamed local variables and struct members accordingly.
Also update a bunch of comments to reflect the renaming and the
accumulated changes that the code has undergone.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since `packed_ref_iterator` is now delegating to
`mmapped_ref_iterator` rather than `cache_ref_iterator` to do the
heavy lifting, there is no need to keep the two iterators separate. So
"inline" `mmapped_ref_iterator` into `packed_ref_iterator`. This
removes a bunch of boilerplate.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now that the `packed-refs` backend doesn't use `ref_cache`, there is
nobody left who might want to store peeled values of references in
`ref_cache`. So remove that feature.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now that everything has been changed to read what it needs directly
out of the `packed-refs` file, `packed_ref_store` doesn't need to
maintain a `ref_cache` at all. So get rid of it.
First of all, this will save a lot of memory and lots of little
allocations. Instead of needing to store complicated parsed data
structures in memory, we just mmap the file (potentially sharing
memory with other processes) and parse only what we need.
Moreover, since the mmapped access to the file reads only the parts of
the file that it needs, this might save reading all of the data from
disk at all (at least if the file starts out sorted).
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We're about to stop storing packed refs in a `ref_cache`. That means
that the only way we have left to optimize `peel_ref()` is by checking
whether the reference being peeled is the one currently being iterated
over (in `current_ref_iter`), and if so, using `ref_iterator_peel()`.
But this can be done generically; it doesn't have to be implemented
per-backend.
So implement `refs_peel_ref()` in `refs.c` and remove the `peel_ref()`
method from the refs API.
This removes the last callers of a couple of functions, so delete
them. More cleanup to come...
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Instead of reading the reference from the `ref_cache`, read it
directly from the mmapped buffer.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now that we have an efficient way to iterate, in order, over the
mmapped contents of the `packed-refs` file, we can use that directly
to implement reference iteration for the `packed_ref_store`, rather
than iterating over the `ref_cache`. This is the next step towards
getting rid of the `ref_cache` entirely.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
It doesn't actually matter now, because the references are only
iterated over to fill the associated `ref_cache`, which itself puts
them in the correct order. But we want to get rid of the `ref_cache`,
so we want to be able to iterate directly over the `packed-refs`
buffer, and then the iteration will need to be ordered correctly.
In fact, we already write the `packed-refs` file sorted, but it is
possible that other Git clients don't get it right. So let's not
assume that a `packed-refs` file is sorted unless it is explicitly
declared to be so via a `sorted` trait in its header line.
If it is *not* declared to be sorted, then scan quickly through the
file to check. If it is found to be out of order, then sort the
records into a new memory-only copy. This checking and sorting is done
quickly, without parsing the full file contents. However, it needs a
little bit of care to avoid reading past the end of the buffer even if
the `packed-refs` file is corrupt.
Since *we* always write the file correctly sorted, include that trait
when we write or rewrite a `packed-refs` file. This means that the
scan described in the previous paragraph should only have to be done
for `packed-refs` files that were written by older versions of the Git
command-line client, or by other clients that haven't yet learned to
write the `sorted` trait.
If `packed-refs` was already sorted, then (if the system allows it) we
can use the mmapped file contents directly. But if the system doesn't
allow a file that is currently mmapped to be replaced using
`rename()`, then it would be bad for us to keep the file mmapped for
any longer than necessary. So, on such systems, always make a copy of
the file contents, either as part of the sorting process, or
afterwards.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Keep a copy of the `packed-refs` file contents in memory for as long
as a `packed_ref_cache` object is in use:
* If the system allows it, keep the `packed-refs` file mmapped.
* If not (either because the system doesn't support `mmap()` at all,
or because a file that is currently mmapped cannot be replaced via
`rename()`), then make a copy of the file's contents in
heap-allocated space, and keep that around instead.
We base the choice of behavior on a new build-time switch,
`MMAP_PREVENTS_DELETE`. By default, this switch is set for Windows
variants.
After this commit, `MMAP_NONE` and `MMAP_TEMPORARY` are still handled
identically. But the next commit will introduce a difference.
This whole change is still pointless, because we only read the
`packed-refs` file contents immediately after instantiating the
`packed_ref_cache`. But that will soon change.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
No code has been changed. This will make subsequent patches more
self-contained.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If a reference is broken, suppress its peeled value.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add a new `mmapped_ref_iterator`, which can iterate over the
references in an mmapped `packed-refs` file directly. Use this
iterator from `read_packed_refs()` to fill the packed refs cache.
Note that we are not yet willing to promise that the new iterator
generates its output in order. That doesn't matter for now, because
the packed refs cache doesn't care what order it is filled.
This change adds a lot of boilerplate without providing any obvious
benefits. The benefits will come soon, when we get rid of the
`ref_cache` for packed references altogether.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Rather than store the peeling state (i.e., the one defined by traits
in the `packed-refs` file header line) in a local variable in
`read_packed_refs()`, store it permanently in `packed_ref_cache`. This
will be needed when we stop reading all packed refs at once.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Instead of copying data from the `packed-refs` file one line at time
and then processing it, process the data in place as much as possible.
Also, instead of processing one line per iteration of the main loop,
process a reference line plus its corresponding peeled line (if
present) together.
Note that this change slightly tightens up the parsing of the
`packed-refs` file. Previously, the parser would have accepted
multiple "peeled" lines for a single reference (ignoring all but the
last one). Now it would reject that.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The old code parsed the traits in the `packed-refs` header by looking
for the string " trait " (i.e., the name of the trait with a space on
either side) in the header line. This is fragile, because if any other
implementation of Git forgets to write the trailing space, the last
trait would silently be ignored (and the error might never be
noticed).
So instead, use `string_list_split_in_place()` to split the traits
into tokens then use `unsorted_string_list_has_string()` to look for
the tokens we are interested in. This means that we can read the
traits correctly even if the header line is missing a trailing
space (or indeed, if it is missing the space after the colon, or if it
has multiple spaces somewhere).
However, older Git clients (and perhaps other Git implementations)
still require the surrounding spaces, so we still have to output the
header with a trailing space.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This tightens up the parsing a bit; previously, stray header-looking
lines would have been processed.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
It's still done in a pretty stupid way, involving more data copying
than necessary. That will improve in future commits.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Extract some helper functions for reporting errors. While we're at it,
prevent them from spewing unlimited output to the terminal. These
functions will soon have more callers.
These functions accept the problematic line as a `(ptr, len)` pair
rather than a NUL-terminated string, and `die_invalid_line()` checks
for an EOL itself, because these calling conventions will be
convenient for future callers. (Efficiency is not a concern here
because these functions are only ever called if the `packed-refs` file
is corrupt.)
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
References are iterated over in order by refname, but reflogs are not.
Some consumers of reference iteration care about the difference. Teach
each `ref_iterator` to keep track of whether its output is ordered.
`overlay_ref_iterator` is one of the picky consumers. Add a sanity
check in `overlay_ref_iterator_begin()` to verify that its inputs are
ordered.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now the outside world interacts with the packed ref store only via the
generic refs API plus a few lock-related functions. This allows us to
delete some functions that are no longer used, thereby completing the
encapsulation of the packed ref store.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Implement `packed_delete_refs()` using a reference transaction. This
means that `files_delete_refs()` can use `refs_delete_refs()` instead
of `repack_without_refs()` to delete any packed references, decreasing
the coupling between the classes.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Implement the methods needed to support reference transactions for
the packed-refs backend. The new methods are not yet used.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The old code incremented the packed ref cache reference count when
acquiring the packed-refs lock, and decremented the count when
releasing the lock. This is unnecessary because:
* Another process cannot change the packed-refs file because it is
locked.
* When we ourselves change the packed-refs file, we do so by first
modifying the packed ref-cache, and then writing the data from the
ref-cache to disk. So the packed ref-cache remains fresh because any
changes that we plan to make to the file are made in the cache first
anyway.
So there is no reason for the cache to become stale.
Moreover, the extra reference count causes a problem if we
intentionally clear the packed refs cache, as we sometimes need to do
if we change the cache in anticipation of writing a change to disk,
but then the write to disk fails. In that case, `packed_refs_unlock()`
would have no easy way to find the cache whose reference count it
needs to decrement.
This whole issue will soon become moot due to upcoming changes that
avoid changing the in-memory cache as part of updating the packed-refs
on disk, but this change makes that transition easier.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The previous commit taught the tempfile code to give up
ownership over tempfiles that have been renamed or deleted.
That makes it possible to use a stack variable like this:
struct tempfile t;
create_tempfile(&t, ...);
...
if (!err)
rename_tempfile(&t, ...);
else
delete_tempfile(&t);
But doing it this way has a high potential for creating
memory errors. The tempfile we pass to create_tempfile()
ends up on a global linked list, and it's not safe for it to
go out of scope until we've called one of those two
deactivation functions.
Imagine that we add an early return from the function that
forgets to call delete_tempfile(). With a static or heap
tempfile variable, the worst case is that the tempfile hangs
around until the program exits (and some functions like
setup_shallow_temporary rely on this intentionally, creating
a tempfile and then leaving it for later cleanup).
But with a stack variable as above, this is a serious memory
error: the variable goes out of scope and may be filled with
garbage by the time the tempfile code looks at it. Let's
see if we can make it harder to get this wrong.
Since many callers need to allocate arbitrary numbers of
tempfiles, we can't rely on static storage as a general
solution. So we need to turn to the heap. We could just ask
all callers to pass us a heap variable, but that puts the
burden on them to call free() at the right time.
Instead, let's have the tempfile code handle the heap
allocation _and_ the deallocation (when the tempfile is
deactivated and removed from the list).
This changes the return value of all of the creation
functions. For the cleanup functions (delete and rename),
we'll add one extra bit of safety: instead of taking a
tempfile pointer, we'll take a pointer-to-pointer and set it
to NULL after freeing the object. This makes it safe to
double-call functions like delete_tempfile(), as the second
call treats the NULL input as a noop. Several callsites
follow this pattern.
The resulting patch does have a fair bit of noise, as each
caller needs to be converted to handle:
1. Storing a pointer instead of the struct itself.
2. Passing the pointer instead of taking the struct
address.
3. Handling a "struct tempfile *" return instead of a file
descriptor.
We could play games to make this less noisy. For example, by
defining the tempfile like this:
struct tempfile {
struct heap_allocated_part_of_tempfile {
int fd;
...etc
} *actual_data;
}
Callers would continue to have a "struct tempfile", and it
would be "active" only when the inner pointer was non-NULL.
But that just makes things more awkward in the long run.
There aren't that many callers, so we can simply bite
the bullet and adjust all of them. And the compiler makes it
easy for us to find them all.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since the lockfile code is based on the tempfile code, it
has some of the same problems, including that close_lock_file()
erases the tempfile's filename buf, making it hard for the
caller to write a good error message.
In practice this comes up less for lockfiles than for
straight tempfiles, since we usually just report the
refname. But there is at least one buggy case in
write_ref_to_lockfile(). Besides, given the coupling between
the lockfile and tempfile modules, it's less confusing if
their close() functions have the same semantics.
Just as the previous commit did for close_tempfile(), let's
teach close_lock_file() and its wrapper close_ref() not to
rollback on error. And just as before, we'll give them new
"gently" names to catch any new callers that are added.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
One of the tricks that `contrib/workdir/git-new-workdir` plays is to
making `packed-refs` in the new workdir a symlink to the `packed-refs`
file in the original repository. Before
42dfa7ecef ("commit_packed_refs(): use a staging file separate from
the lockfile", 2017-06-23), a lockfile was used as the staging file,
and because the `LOCK_NO_DEREF` was not used, the pointed-to file was
locked and modified.
But after that commit, the staging file was created using a tempfile,
with the end result that rewriting the `packed-refs` file in the
workdir overwrote the symlink rather than the original `packed-refs`
file.
Change `commit_packed_refs()` to use `get_locked_file_path()` to find
the path of the file that it should overwrite. Since that path was
properly resolved when the lockfile was created, this restores the
pre-42dfa7ecef behavior.
Also add a test case to document this use case and prevent a
regression like this from recurring.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Reviewed-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Ævar Arnfjörð Bjarmason