We prefer matches closer to the middle to balance recursion, as introduced in
d3deb406b55b.
For ranges with uneven length, matches starting exactly in the middle should
have preference. That will be optimal for matches of length 1. We will thus
accept equality in the half check.
For ranges with even length, half was ceil'ed when calculated but we got the
preference for low matches from the 'less than half' check. To get the same
result as before when we also accept equality, floor it. Without that,
test-annotate.t would show some different (still correct but less optimal)
results.
This will change the heuristics. Tests shows a slightly different output - and
sometimes slightly smaller bundles.
The bundle size for 4.0 (hg bundle --base null -r 4.0 x.hg) happens to go from
22804885 to 22803824 bytes - an 0.005% reduction.
This broke in c7236da49964 due to a refactored manifest API.
The fix is a bit hacky - perfbdiff doesn't yet support tree manifests
for example. But it gets the job done.
A test has been added for --alldata so this doesn't happen again.
Since we now wait child processes in non-blocking way (changed by 6c7588a50638
and 13c3aefdee29), we don't have to kill them in the middle of the waitpid()
loop. This change will help solving a possible race of waitpid()-pids.discard()
sequence and another SIGCHLD.
waitforworkers() is called by cleanup(), in which case we do killworkers()
beforehand so we can remove killworkers() from waitforworkers().
killworkers() iterates over pids, which can be updated by SIGCHLD handler.
So we should either copy pids or prevent killworkers() from being interrupted
by SIGCHLD. I chose the latter as it is simpler and can make pids handling
more consistent.
This fixes a possible "set changed size during iteration" error at
killworkers() before cleanup().
We've only got one kind of pyflakes failure left in our codebase, so
it's time to switch over to a blacklist-based checking scheme. I've
left in the filtering of two undefined names for now out of paranoia,
but those can probably go too.
We do this so that any linters installed via pip install --user don't
break. See https://docs.python.org/2/library/site.html#site.USER_BASE
for a description of what this nonsense is all about.
An alternative would be to not set HOME, but that'll cause other
problems (see issue2707), or to forward every single path entry from
sys.path in PYTHONPATH (which seems sketchy in its own way).
We have performance concerns on "iterfile" as it is 4X slower on normal
files. While modern systems have the nice property that reading a "fast"
(on-disk) file cannot be interrupted and should be made use of.
This patch dumps the related knowledge in comments. And "iterfile" chooses
code paths wisely:
1. If it's CPython 3, or PyPY, use the fast path.
2. If fp is a normal file, use the fast path.
3. If fp is not a normal file and CPython version >= 2.7.4, use the same
workaround (4x slower) as before.
4. If fp is not a normal file and CPython version < 2.7.4, use another
workaround (2x slower but may block longer then necessary) which
basically re-invents the buffer + readline logic in Python.
This will give us good confidence on both correctness and performance
dealing with EINTR in iterfile(fp) for all known supported Python versions.
According to POSIX specification, just having group write access to a
file causes EPERM at invocation of os.utime() with an explicit time
information (e.g. working on the repository shared by group access
permission).
To ignore EPERM at closing file object in such case, this patch makes
checkambigatclosing._checkambig() use filestat.avoidambig() introduced
by previous patch.
Some functions below imply this code path at truncation of an existing
(= might be owned by another user) file.
- strip() in repair.py, introduced by 4d0a08431b6f
- _playback() in transaction.py, introduced by 48fe04792102
This is a variant of issue5418.
According to POSIX specification, just having group write access to a
file causes EPERM at invocation of os.utime() with an explicit time
information (e.g. working on the repository shared by group access
permission).
To ignore EPERM at renaming in such case, this patch makes
vfs.rename() use filestat.avoidambig() introduced by previous patch.
Now, advancing stat.st_mtime by os.utime() is used to avoid file stat
ambiguity. But according to POSIX specification, utime(2) with an
explicit time information is permitted only for a process with:
- the effective user ID equal to the user ID of the file, or
- appropriate privileges
http://pubs.opengroup.org/onlinepubs/9699919799/functions/utime.html
Therefore, just having group write access to a file causes EPERM at
applying os.utime() on it (e.g. working on the repository shared by
group access permission).
This patch adds class filestat utility function avoidamgig() to avoid
file stat ambiguity but skip it if EPERM.
It is reasonable to always ignore EPERM, because utime(2) causes EPERM
only in the case described above (EACCES is used only for utime(2)
with NULL).
Now that we have a SIGCHLD hander, and it could get executed when waiting
for I/O. It's no longer necessary to have a separated waitpid thread. So
just remove it.
As planned by previous patches, add a SIGCHLD handler to get notifications
about worker exits, and deals with worker failure immediately.
Note that the SIGCHLD handler gets unregistered before killworkers(), so
SIGCHLD won't interrupt "killworkers" - making it harder to send kill
signals to waited processes.
We are going to use it in the SIGCHLD handler. The handler will be executed
in the main thread with the non-blocking version of waitpid, while the
waitforworkers thread runs the blocking version. It's possible that one of
them collects a worker and makes the other error out (no child to wait).
This patch handles these errors: ECHILD is ignored. EINTR needs a retry.
The "pids" set is designed to be only modifiable by "waitforworkers". And we
only remove items after a successful waitpid. Since a child process can only
be "waitpid"-ed once. It's guaranteed that "pids.remove(p)" won't be called
with duplicated "p"s. And once a "p" is removed from "pids", that "p" does
not need to be killed or waited any more.
43e3fb1c484e introduced a call to fctx.parents() for each line in
annotate output. This function call isn't cheap, as it requires
linkrev adjustment.
Since multiple lines in annotate output tend to belong to the same
file revision, a cache of fctx.parents() lookups for each input
should be effective in the common case. So we implement one.
Since the cache has to precompute parents so an aborted generator
doesn't leave an incomplete cache, we could just return a list.
However, we preserve the generator for backwards compatibility.
The effect of this change when requesting /annotate/96ca0ecdcfa/
browser/locales/en-US/chrome/browser/downloads/downloads.dtd on
the mozilla-aurora repo is significant:
p1(43e3fb1c484e) 5.5s
43e3fb1c484e: 66.3s
this patch: 10.8s
We're still slower than before. But only by ~2x instead of ~12x.
On the tip revisions of layout/base/nsCSSFrameConstructor.cpp file in
the mozilla-unified repo, time went from 12.5s to 14.5s and back to
12.5s. I'm not sure why the mozilla-aurora repo is so slow.
Looking at the code of basefilectx.parents(), there is room for
further improvements. Notably, we still perform redundant calls to
filelog.renamed() and basefilectx._parentfilectx(). And
basefilectx.annotate() also makes similar calls, so there is potential
for object reuse. However, introducing caches here are not appropriate
for the stable branch.
Currently the warning is ambiguous about whether the new tag (possibly specified
via --rev) is being added on a branch head or whether the working directory is
based on a branch head. Clarify the error message to eliminate this ambiguity.
Now, all URL in Mercurial source tree should refer mercurial-scm.org
domain instead of selenic.com.
*.po files are ignored in this patch, because they might contain
msgid/msgstr coming from old source files.
This ignorance seems safe enough, because such msgstr should be
ignored at runtime, because:
- msgid corresponded to it should be invalid, or
- msgstr itself should be marked as fuzzy at synchronized to recent hg.pot
If any additional examination for *.po files is needed in the future,
let i18n/check-translation.py achieve such examination.
BTW, some binary files (e.g. *.png) are meaningless for checking
reference to old domain in this patch, but aren't ignored like as *.po
files, because excluding multiple suffixes is difficult for regexp
matching.
Before this patch, check-code.py applies filtering on the file
content, to which filtering of previous check is already applied.
This might hide issues, which should be detected by a subsequent check
in "checks" list.
Fortunately, this problem hasn't appeared, because there is no
overlapping of filename matching (examined in the order below).
1. *.py or *.cgi
2. test-* (not *.t suffix)
3. *.c or *.h
4. *.t
5. *.txt
6. *.tmpl
For example, adding a test, which wants to examine raw comment text in
*.py files, at the end of current "checks" list doesn't work as
expected, because a filter for *.py files normalizes comment text in
them.
Putting such test at the beginning of "checks" list also resolves this
problem, but such dependence on the order decreases maintainability of
check-code.py itself.
This patch discards filtering result of previous check at the
beginning of each checks, for independence of each checks.
Some errors could in some cases show unfortunate scary and confusing warnings
from the httppeer delstructors:
abort: nodename nor servname provided, or not known
Exception AttributeError: "'httpspeer' object has no attribute 'urlopener'" in <bound method httpspeer.__del__ of <mercurial.httppeer.httpspeer object at 0x106e1f5d0>> ignored```
To mute that, take 8bdb0bb8e209 to the next level and use getattr in __del__.
test-largefiles-update.t creates temporary file exec-bit.patch inside
the working directory for no-execbit platform specific test, but
subsequent tests aren't aware of it.
On execbit platform, subsequent tests can run successfully, because
exec-bit.patch isn't created.
But on no-execbit platform, this temporary file makes subsequent tests
show "? exec-bit.patch" at each "hg status".
journal extension uses util.shellquote() to record command line, but
result of it depends on runtime platform: double quotation is used on
Windows and OpenVMS, but single quotation is used otherwise.
test-journal-share.t sometimes specifies commit messages including
white space on command line. It makes journal output depend on runtime
platform, but commit message itself isn't important in this test case.
On Windows, strftime() doesn't support format code "%s", and it causes
"invalid format string" error.
https://msdn.microsoft.com/en-us/library/fe06s4ak.aspx
test-command-template.t examines not seconds value in UTC, but
arithmetic calculation. Therefore, using format code "%Y" instead of
"%s" should be reasonable.
FYI:
- Python standard library reference doesn't list "%s" up in format
code list required for "C standard (1989 version)", even though it
also mentions that additional format codes are required for "C
standard (1999 version)"
https://docs.python.org/2.7/library/datetime.html#strftime-and-strptime-behavior
- The Open Group Base Specifications Issue 7 (IEEE Std 1003.1-2008,
2016 Edition) doesn't require strftime to support format code "%s"
http://pubs.opengroup.org/onlinepubs/9699919799/functions/strftime.html
- "man strftime" of (Open/Oracle) Solaris and Mac OS X (= UNIX
certified OSs) describes about format code "%s"
If environment variable looks like PATH or so (e.g. any of components
joined by ":" contains "/"), ":" in it is replaced with ";" by MinGW
at spawning Windows native process, to follow path concatenation style
of Windows.
Therefore, "bundle:../full.hg" is converted into "bundle;..\full.hg"
on MinGW.
Difference between "/" and "\" is automatically ignored by "(glob)",
but difference between ":" and ";" should be globed explicitly.
On Windows platform, invoking printenv.py directly via hook is
problematic, because:
- unless binding between *.py suffix and python runtime, application
selector dialog is displayed, and running test is blocked at each
printenv.py invocations
- it isn't safe to assume binding between *.py suffix and python
runtime, because application binding is easily broken
For example, installing IDE (VisualStudio with Python Tools, or
so) often requires binding between source files and IDE itself.
This patch invokes printenv.py via sh -c for test portability. This is
a kind of follow up for 9e4331825bea, which eliminated explicit
"python" for printenv.py. There are already other 'sh -c "printenv.py"'
in *.t files, and this fix should be reasonable.
This changes were confirmed in cases below:
- without any application binding for *.py suffix
- with binding between *.py suffix and VisualStudio
This patch also replaces "echo + redirection" style with "heredoc"
style, because:
- hook command line is parsed by cmd.exe as shell at first, and
- single quotation can't quote arguments on cmd.exe, therefore,
- "printenv.py foobar" should be quoted by double quotation, but
- nested quoting (or tricky escaping) isn't readable
I somehow ended up in a situation where hg crashed on an unlink I introduced in
8fd3fc1ef4c6.
I don't know how it happened and can't reproduce it. It seems like it only can
happen when the file is removed between the time of check in a working
directory context walk that finds a standin file, and the time of use when we
try to remove it because the corresponding largefile doesn't exist.
But better safe than sorry: replace the plain unlink with unlinkpath with
ignoremissing=True. That will also remove remaining empty directories, which
arguably is more correct.
This is a fix for a regression introduced by the patches for issue4028.
The test changes are due to us doing fewer _checkcopies searches now, which
makes some test outputs revert to the pre-issue4028 behavior. That issue itself
remains fixed, we only skip copy tracing for files where it isn't relevant.
As a nice side effect, this makes copy detection much faster when tracing
backwards through lots of renames.
cl._partialmatch() can be pretty slow if hidden revisions are involved. This
patch cancels the slowdown introduced by the previous patch by using an
unfiltered changelog, which means shortest(node) isn't always the shortest.
The result isn't perfect, but seems okay as long as shortest(node) is short
enough to type and can be used as an identifier.
(with hidden revisions)
% hg log -R hg-committed -r0:20000 -T '{node|shortest}\n' --time > /dev/null
(.^^) time: real 1.530 secs (user 1.480+0.000 sys 0.040+0.000)
(.^) time: real 43.080 secs (user 43.060+0.000 sys 0.030+0.000)
(.) time: real 1.680 secs (user 1.650+0.000 sys 0.020+0.000)
cl.index.partialmatch() isn't a drop-in replacement for cl._partialmatch().
It has no knowledge about hidden revisions, and it raises ValueError if a node
shorter than 4 chars is given. Instead, use index.partialmatch() through
cl._partialmatch(), which has no such problems and gives the identical result
with/without --pure.
The test output was sampled with --pure without this patch, which shows the
most correct result. However, we'll need to switch to using an unfiltered
changelog because _partialmatch() of a filtered changelog can be an order of
magnitude slower.
(with hidden revisions)
% hg log -R hg-committed -r0:20000 -T '{node|shortest}\n' --time > /dev/null
(.^) time: real 1.530 secs (user 1.480+0.000 sys 0.040+0.000)
(.) time: real 43.080 secs (user 43.060+0.000 sys 0.030+0.000)
On some platforms, cwd can't be removed. In which case, util.unlinkpath()
continues with no error since the failure of directory removal isn't critical.
So it doesn't make sense to run the test added by 6395630fdfdc on those
platforms. OTOH, we need to run the test in test-rebase-scenario-global.t
since the repository is referenced after that.
Certifi is currently incompatible with py2exe; the Python code for certifi gets
included in library.zip, but not the cacert.pem file - and even if it were
included, SSLContext can't load a cacert.pem file from library.zip.
This currently makes it impossible to build a standalone Windows version of
Mercurial.
Guard against this, and possibly other situations where a module with the name
"certifi" exists, but is not usable.
There was a "leak", apparently introduced in b37a67b41690. When running:
hg = hglib.open('repo')
while True:
hg.log("max(branch('default'))")
all filteredset instances from branch() would be cached indefinitely by the
@util.cachefunc annotation on the max() implementation.
util.cachefunc seems dangerous as method decorator and is barely used elsewhere
in the code base. Instead, just open code caching by having the min/max
methods replace themselves with a plain lambda returning the result.
Over the past week I've had to instruct multiple people to run
Python code to query the ssl module to see what TLS protocol support
is present. I think it would be useful for `hg debuginstall` to print
this info to make it easier to access and debug why Mercurial is
complaining about using an insecure TLS 1.0 protocol.
Ideally we'd also print the path to the CA cert bundle. But the APIs
for querying that in sslutil can emit warnings, making it slightly
more difficult to integrate into `hg debuginstall`. That work will
have to wait for another day.
Same as in the last commit, the old treemanifestctx stored a reference to the
revlog. If the inmemory revlog became invalid, the ctx now held an old copy and
would be incorrect. To fix this, we need the ctx to go through the manifestlog
for each access.
This is the same pattern that changectx already uses (it stores the repo, and
accesses commit data through self._repo.changelog).
The old manifestctx stored a reference to the revlog. If the inmemory revlog
became invalid, the ctx now held an old copy and would be incorrect. To fix
this, we need the ctx to go through the manifestlog for each access.
This is the same pattern that changectx already uses (it stores the repo, and
accesses commit data through self._repo.changelog).