"hg status" may treat cache missed largefiles as "removed" incorrectly.
assumptions for problem case:
- there is no cache for largefile "L"
- at first, update working directory to the revision in which "L" is
not yet added,
- then, update working directory to the revision in which "L" is
already added
and now, "hg status" treats "L" as "removed".
current implementation does not allocate entry for cache missed
largefile in ".hg/largefiles/dirstate", but files without
".hg/largefiles/dirstate" entry are treated as "removed" by largefiles
extension.
"hg revert" can not recover from this situation, but "rm -rf
.hg/largefiles", because it causes dirstate rebuilding.
this patch invokes normallookup() for cache missed largefiles to
allocate entry in ".hg/largefiles/dirstate", so "hg status" can treat
it as "missing" correctly.
When storing/restoring a nullmerge (-2), a 'standard' conversion was made
and an existing changeset was wrongly used.
Nullmerge should instead be treated as a special case.
Prior to this patch "hg diff -U0", i.e., zero lines of context, would
output hunk headers with a start line one greater than what GNU patch
and git output. Guido van Rossum documents the unified diff format[1]
as having a start line value "one lower than one would expect" for
zero length hunks.
Comparing the behaviour of the three systems prior to this patch in
transforming
c1
c3
to
c1
c2
c3
- GNU "diff -U0" reports the hunk as "@@ -1,0 +2 @@"
- "git diff -U0" reports the hunk as "@@ -1,0 +2 @@"
- "hg diff -U0" reports the hunk as "@@ -2,0 +2,1 @@"
After this patch, "hg diff -U0" reports "@@ -1,0 +2,1 @@".
Since "hg export --config diff.unified=0" outputs zero-context unified
diffs, "hg import" has also been updated to account for start lines
one less than expected for zero length hunk ranges.
[1]: http://www.artima.com/weblogs/viewpost.jsp?thread=164293
With renames like:
a -> b
a/c -> a/c
We were ignoring or duplicating the second one instead of leaving files
unchanged or moving them to their proper destination only.
To avoid this, we process the files in reverse lexicographic order, from most
to least specific change, and ignore files already processed.
v2:
- Add a test
- Change "reverse=1" into "reverse=True"
This fix mirrors the changes made to test-doctest.py in 04cfbbc5ae97
and 39599b7929c4.
Without this change, tests running heredoctest.py can fail on certain
versions of OS X when TERM is set to xterm-256color:
$ /opt/local/Library/Frameworks/Python.framework/Versions/2.7/Resources/Python.app/Contents/MacOS/Python -m heredoctest <<EOF
> >>> open('b', 'w').write('this' * 1000)
> EOF
+ \x1b[?1034h (no-eol) (esc)
A similar problem occurs with test-url.py:
$ ./run-tests.py test-url.py
--- .../tests/test-url.py.out
+++ .../tests/test-url.py.err
@@ -0,0 +1 @@
+
ERROR: .../test-url.py output changed
!
Failed test-url.py: output changed
# Ran 1 tests, 0 skipped, 1 failed.
# this change redones part of e0051068893a, backed out by 38c00c035629
Some character encodings use ASCII characters other than
control/alphabet/digit as a part of multi-bytes characters, so direct
replacing with such characters on strings in local encoding causes
invalid byte sequences.
[mpm: test changed to simple doctest]
$DAEMON_PIDS is used by tests to make sure there will be no leftover
processes, the cycling through ports is needed because they are not
available for a new bind that quickly on some systems.
There are not enough $HGPORT variables available for each hg serve,
so use the killdaemons script before reusing ports.
While Chrome, Firefox, and IE 6+ support the current date format being
passed to Date(), Safari doesn't:
> new Date('Mon Oct 24 13:58:01 2011 +0200')
Invalid Date
However, the rfc822date format--officially supported by
ECMAScript[1]--does work:
> new Date('Mon, 24 Oct 2011 13:58:01 +0200')
Mon Oct 24 2011 04:58:01 GMT-0700 (PDT)
This change replaces all instances of {date|date} in HTML with
{date|rfc822date}. For elements that only have the "age" class,
there's no outward change for users with JavaScript enabled. For
elements with both the "age" and "date" classes, the full date
displayed uses the new format.
Tested in IE 6, Safari 5.1.1, Google Chrome 15, and Firefox 7.0.1.
[1]: https://developer.mozilla.org/en/JavaScript/Reference/Global_Objects/Date/parse
- old-style patterns without ^ were getting improperly anchored
- finditer was matching against beginning of line poorly
- \s was matching newlines
- [^x] was matching newlines
so we:
- remove earlier hacks for multiline matching
- fix unified test anchoring by adding .*
- replace \s with [ \t]
- replace [^x] with [^\nx]
- force all matches into multiline mode so ^ anchors work
This uncovers a number of test issues that are then repaired.
The code was using the size of a symlink's target, thus wrongly making symlinks
to large files into largefiles themselves. This can be demonstrated by
deleting the symlink and then doing an 'hg up' or 'hg up -C' to restore the
symlink.
This reveals a small bug: revert reports "reverting .hglf/largefile"
when it really should report "reverting largefile". I don't see an
easy fix, though (short of using ui.pushbuffer() to interfere with
revert's output).
The 'Bin' marker was added to every changed file for which we could not find
any diff changes. This included binary files but also copy/renames and mode
changes. Since Mercurial regular diff format emits a 'Binary file XXX has
changed' line when fed with binary files, we use that and the usual git marker
to tell them from other cases. In particular, new empty files are no longer
reported as binary.
Still, this fix is not complete since copy/renames/mode changes are now
reported as '0' lines changes, instead of 'Bin'.
On Windows, we store symlinks as plain files with the link contents.
Via user error or NFS/Samba assistance, these files often end up with
'normal' file contents. Committing these changes thus gives an
invalid symlink that can't be checked out on Unix.
Here we filter out any modified symlink placeholders that look
suspicious when computing status:
- more than 1K (looks more like a normal file)
- contain NULs (not allowed on Unix, probably a binary)
- contains \n (filenames can't contain \n, very unusual for symlinks,
very common for files)
This gets us very close to full test coverage of lfconvert. The only
features not tested are:
- .hgtags conversion
- abort when largefile becomes symlink
- instead of converting a single changeset with a single file, convert
two changesets which each add a mix of normal and large files
- make one of the large files not-so-large, but identified as large
by filename
- use shorter, simpler filenames
- put some files in a subdirectory
- test that file contents are correctly converted
- drop unnecessary cleanup step
test-largefiles.t is getting pretty big, and there's lots more to test
in lfconvert. And lfconvert is a fairly orthogonal feature to the
rest of largefiles.
Before this patch, Windows always did the wrong thing with exec bits
when committing a merge: consult the flags in first parent.
Now we manually recompute the result of merging flags at commit time,
which almost always does the right thing (except when there are
conflicts between symlink and exec flags).
To do this, we:
- pull flag synthesis out into its own function
- delay building this function unless it's needed
- add a merge case that compares flags in local and other against the ancestor
This has been tested in multiple ways on Linux:
- running the whole test suite with both old and new code in place,
checking for differences in each flags() result
- running the whole test suite while comparing real on-disk flags
against synthetic ones for merges
- test-issue1802 (from Martin Geisler) which disables exec bit
checking on Unix
This reintroduces tests for the case described by issue3066, among others:
- vanilla clients and servers should still be able to talk to largefiles
clients and servers about vanilla repos
- vanilla clients should not attempt to clone largefiles repos over http and
should report a useful error message (issue3066; this appears to work
correctly)
- vanilla clients should not attempt to clone largefiles repos over ssh, again
with a useful error message
- largefiles clients should not attempt to push largefile changes to vanilla
servers
When renaming a patch A as B where B was previously qfolded into A and
therefore marked as removed, a versioned MQ would first restore B before
marking it as a copy of A, thus losing A changes. The undelete() call is
probably a left-over, wctx.copy() explicitely handles the case where the
destination is removed.
Also note that status command represents "hg rm b; hg mv a b" as:
A b
a
R a
which explains the first hunk in test-mq-qrename.t.
This changeset flips the default value of ui.commitsubrepos setting
from True to False and adds a --subrepos flag to commit.
The commit, status, and diff commands behave like this with regard to
recusion and the ui.commitsubrepos setting:
| recurses | recurses
| by default | with --subrepos
--------+---------------+----------------
commit: | commitsubrepo | True
status: | False | True
diff: | False | True
By changing the default from True to False, the table becomes
consistent in the two columns:
* without --subrepos on the command line, commit will abort if a
subrepo is dirty and status/diff wont show changes inside subrepos.
* with --subrepos, all three commands will recurse.
A --subrepos flag on the command line overrides the config settin.g
Matching lines without trailing '\n' was missing the last character.
That seems to have been an unintended side effect of 8abe3f27975c.
The test in c21748e4cd4d documents the bad behaviour.
Before, we deleted foo when we determined that there were zero
changesets in the foo subrepo. Any files in foo was deleted too. We
now delete foo/.hg instead, and the normal checks in hg.clone will
then abort if there are untracked files in foo.
Before, running 'hg archive -S' could result in
abort: unknown revision '65903cebad86f1a84bd4f1134f62fa7dcb7a1c98'!
if a subrepo was missing completely or had missing changesets. Now,
the missing changesets will be pulled or cloned as appropriate.
This make Mercurial subrepos match Git subrepos which already took
care to fetch any missing commits before starting the archive.
This option allow a strict set of revision to be specified instead of using -s
or -b. Rebase will refuse start if striping rebased changeset will strip non
rebased changeset. Rebase will refuse to work on set with multiple root.
The buildstate function now take a set of revs. Logic related to --source and
--base option have been moved in the main rebase function.
In the process this fixes a bug where the wrong source changeset might be pick.
This explain the changes in hgext/rebase.py
hgrc(5) already implies that this works, so we might as well support it.
Another approach would be to implement this in util.findexe(): that
would benefit other callers of findexe(), e.g. convert and anyone
calling the user's editor. But findexe() is really implemented in
both posix.py and windows.py, so this would make both of those modules
depend on util.py: not good. So keep it narrow and only for merge
tools.
Older publish=True was:
1) Content of Publishing server are seen as public by client.
2) Any changegroup *added* to a publish=True server is public.
New definition are:
1) Content of Publishing server are seen as public by client.
2) Any changegroup *pushed* to a publish=True server is public.
See mercurial/phase.py documentation for exact final behavior
Change the behavior so that the sender (the "From" header in the notification
mail) in case of the "changegroup" hook is the user that did the first commit
in the changegroup. The option is configurable, if you set "notify.fromauthor"
to "True" in your config, the new behavior is activated. If you do not set the
option, the behavior is as before. The commit adds to an existing test to show
various aspects of the changed behavior.
What is a "publishing repository"?
==================================
Setting a repository as "publishing" alter its behavior **when used as a
server**: all changesets are **seen** as public changesets by clients.
So, pushing to a "publishing" repository is the most common way to make
changesets public: pushed changesets are seen as public on the remote side and
marked as such on local side.
Note: the "publishing" property have no effects for local operations.
Old repository are publishing
=============================
Phase is the first step of a series of features aiming at handling mutable
history within mercurial. Old client do not support such feature and are unable
to hold phase data. The safest solution is to consider as public any changeset
going through an old client.
Moreover, most hosting solution will not support phase from the beginning.
Having old clients seen as public repositories will not change their usage:
public repositories where you push *immutable* public changesets *shared* with
others.
Why is "publishing" the default?
================================
We discussed above that any changeset from a non-phase aware repository should
be seen as public. This means that in the following scenario, X is pulled as
public::
~/A$ old-hg init
~/A$ echo 'babar' > jungle
~/A$ old-hg commit -mA 'X'
~/A$ cd ../B
~/B$ new-hg pull ../A # let's pretend A is served by old-hg
~/B$ new-hg log -r tip
summary: X
phase: public
We want to keep this behavior while creating/serving the A repository with
``new-hg``. Although committing with any ``new-hg`` creates a draft changeset.
To stay backward compatible, the pull must see the new commit as public.
Non-publishing server will advertise them as draft. Having publishing repository
the default is thus necessary to ensure this backward compatibility.
This default value can also be expressed with the following sentence: "By
default, without any configuration, everything you exchange with the outside is
immutable.". This behaviour seems sane.
Why allow draft changeset in publishing repository
=====================================================
Note: The publish option is aimed at controlling the behavior of *server*.
Changeset in any state on a publishing server will **always*** be seen as public
by other client. "Passive" repository which are only used as server for pull and
push operation are not "affected" by this section.
As in the choice for default, the main reason to allow draft changeset in
publishing server is backward compatibility. With an old client, the following
scenario is valid::
~/A$ old-hg init
~/A$ echo 'babar' > jungle
~/A$ old-hg commit -mA 'X'
~/A$ old-hg qimport -r . # or any other mutable operation on X
If the default is publishing and new commits in such repository are "public" The
following operation will be denied as X will be an **immutable** public
changeset. However as other clients see X as public, any pull//push (or event
pull//pull) will mark X as public in repo A.
Allowing enforcement of public changeset only repository through config is
probably something to do. This could be done with another "strict" option or a
third value config for phase related option (mode=public, publishing(default),
mutable)
Makes the 'nothing to merge' abort messages in commands.py consistent with
those in merge.py. Also makes commands.merge() and merge.update() use hints.
The tests show the changes.
check-code has fine warning checks, but they are a bit noisy and nobody used
them.
Now the warnings will be run in the test suite, where a list of accepted
warnings will be maintained.
Those who introduce or touch a line with a warning will now have to update the
whitelist ... or fix the warning.
splitblock() was added to handle blocks returned by bdiff.blocks() which differ
only by blank lines but are not made only of blank lines. I do not know exactly
how it could happen but mdiff.blocks() threshold behaviour makes me think it
can if those blocks are made of very popular lines mixed with popular blank
lines. If it is proven to be wrong, the function can be dropped.
The first implementation made annotate share diff configuration entries. But it
looks like users will user -w/b for annotate but not for diff, on both the
command line and hgweb. Since the latter cannot use command line entries, we
introduce a new [annotate] section duplicating the diff whitespace options.
Change the behavior of the forget command such that explicit paths in
subrepos are handled by forgetting the file in the subrepo. This eliminates the
previous behavior where if you called "hg forget" for an explicit path in a
subrepo, it would state that the file is already untracked.
Globbing is usually used for filenames, so on windows it is reasonable and very
convenient that glob patterns accepts '\' or '/' when the pattern specifies
'/'.
debugstate would always report files as mode 666 or 777 on Windows.
umask is not used on Windows, but faking and using a defalt value of 022
matches what the test suite uses on Unix.
Normally changes in tests are reported like this in diffs:
$ cat foo
- a
+ b
Using -i mode lets us update tests when the new results are correct
and/or populate tests with their output.
But with the standard doctest framework, inline Python sections in
tests changes instead result in a big failure report that's unhelpful.
So here, we replace the doctest calls with a simple compile/eval loop.
Change the behavior of the add command such that explicit paths in
subrepos are always added. This eliminates the previous behavior
where if you called "hg add" for an explicit path in a subrepo
without specifying the -S option, it would be silently ignored.
If you specify patterns, or no arguments at all, the -S option
will still be needed to activate recursion into subrepos.
Add test coverage for the existing behavior where adds of explicit paths in
subrepos are silently ignored. This is in preparation for changing the
behavior of the add command to have better support for subrepos.
Most of the code paths in mq would always pass patch specifications as a
string. Patches can be specified by their index, but one code path passed that
(through pop) to lookup as an integer - all other code paths used a string.
Unfortunately pop and lookup (like many other parts of mq) used the boolean
value of the patch specification to see if it was None, and they would thus
incorrectly handle patch 0 as None.
This patch makes the code comply with the actual internal duck typing of patch
specifications: patch indices must be encoded as strings. The (now) unused code
for partial and thus incorrect handling of indices as integers is removed.
- tweak wording of some error messages
- use consistent capitalization
- always say 'largefile', not 'lfile'
- fix I18N problems
- only raise Abort for errors the user can do something about
Some errors could leave self.urlopener uninitialized and thus cause strange
crashes in __del__.
This member variable is now "declared statically" and checked for assignment
before use.
This adds doctest like syntax to .t files, that can be interleaved with regular
shell code:
$ echo -n a > file
>>> print open('file').read()
a
>>> open('file', 'a').write('b')
$ cat file
ab
The syntax is exactly the same as regular doctests, so multiline statements
look like this:
>>> for i in range(3):
... print i
0
1
2
Each block has its own context, i.e.:
>>> x = 0
>>> print x
0
$ echo 'foo'
foo
>>> print x
will result in a NameError.
Errors are displayed in standard doctest format:
>>> print 'foo'
bar
--- /home/idan/dev/hg/default/tests/test-test.t
+++ /home/idan/dev/hg/default/tests/test-test.t.err
@@ -2,3 +2,16 @@
> >>> print 'foo'
> bar
> EOF
+ **********************************************************************
+ File "/tmp/tmps8X_0ohg-tst", line 1, in tmps8X_0ohg-tst
+ Failed example:
+ print 'foo'
+ Expected:
+ bar
+ Got:
+ foo
+ **********************************************************************
+ 1 items had failures:
+ 1 of 1 in tmps8X_0ohg-tst
+ ***Test Failed*** 1 failures.
+ [1]
As for the implementation, it's quite simple: when the test runner sees a line
starting with '>>>' it converts it, and all subsequent lines until the next
line that begins with '$' to a 'python -m heredoctest <<EOF' call with the
proper heredoc to follow. So if we have this test file:
>>> for c in 'abcd':
... print c
a
b
c
d
$ echo foo
foo
It gets converted to:
$ python -m heredoctest <<EOF
> >>> for c in 'abcd':
> ... print c
> a
> b
> c
> d
> EOF
$ echo foo
foo
And then processed like every other test file by converting it to a sh script.
Writes stdin to a temp file and doctests it.
In the future we might want to spare the temp file and directly call into
doctest.
Also, with some tweaking it seems possible to adjust the line numbers reported
in an error report so they match the ones in the original file.
Previously, if you set an alias for "ci", it'd also shadow "commit"
even though you didn't specify that. This occurred for all commands
with explicit short variations.
This has never worked usefully:
- it can't undo a completely unwanted merge, as it leaves the merge in the DAG
- it can't undo a faulty merge as that means doing a merge correctly,
not simply reverting to one or the other parent
Both of these kinds of merge also require coordinated action among
developers to avoid the bad merge continuing to affect future merges,
so we should stop pretending that backout is of any help here.
As backing out a merge now requires a hidden option, it can't be done
by accident, but will continue to 'work' for anyone who's already
dependent on --parent for some unknown reason.
Now 'rollback' after 'import' is less surprising: it rolls back all of
the imported changesets, not just the last one. As an extra added
benefit, you don't need 'rollback -f' after 'import --bypass', which
was an undesired side effect of fixing issue2998 (f9f52d71c33b)..
Note that this is a different take on issue963, which complained that
rollback after importing multiple patches returned the working dir
parent to the starting point, not to the second-last patch applied.
Since we now rollback the entire import, returning the working dir to
the starting point is entirely logical. So this change also undoes
b12d79024900, the fix to issue963, and updates its tests accordingly.
Bottom line: rollback after import was weird before issue963,
understandable since the fix for issue963, and even better now.
The old code printed (with ui.status()) the changeset ID created by
patch N after committing patch N+1, e.g.
applying patch1
applying patch2
applied 1d4bd90af0e4
where 1d4bd90af0e4 is the changeset ID resulting from patch1. That's
just weird. It's also inconsistent: we only reported the changeset ID
when applying >1 patches. And it's inconsistent with 'commit', which
only tells you the new changeset ID in verbose mode. Finally, the
existing code was I18N-hostile, since it concatenated translated
strings.
The new way is to print the just-created changeset ID with ui.note()
immediately after committing it. It also clarifies what the user
message is for easier I18N.
You can get into trouble if you commit, update back to an older
changeset, and then rollback. The update removes your valuable changes
from the working dir, then rollback removes them history. Oops: you've
just irretrievably lost data running nothing but core Mercurial
commands. (More subtly: rollback from a shared clone that was already
at an older changeset -- no update required, just rollback from the
wrong directory.)
The fix assumes that only "commit" transactions have irreplaceable
data, and allows rolling back non-commit transactions as always. But
when rolling back a commit, check that the working dir is checked out
to tip, i.e. the changeset we're about to destroy. If not, abort. You
can get back the old (dangerous) behaviour with --force.
This matches our pre-existing behavior from:
changeset: 12197:0a3b85866451
user: Christian Ebert <blacktrash@gmx.net>
files: hgext/patchbomb.py tests/test-patchbomb.t
description:
patchbomb: show prompt and selection in non-interactive mode
changeset: 8940:023d1310d8a4
user: Mads Kiilerich <mads@kiilerich.com>
date: Sun Jun 21 03:13:38 2009 +0200
files: mercurial/ui.py tests/test-merge-prompt.out tests/test-merge-tools.out
description:
ui.prompt: Show prompt and selection in non-interactive mode