config: use utf8_bom[] from utf.[ch] in git_parse_source()
Because the function reads one character at the time, unfortunately
we cannot use the easier skip_utf8_bom() helper, but at least we do
not have to duplicate the constant string this way.
With the recent change to ignore the UTF8 BOM at the beginning of
.gitignore files, we now have two codepaths that do such a skipping
(the other one is for reading the configuration files).
Introduce utf8_bom[] constant string and skip_utf8_bom() helper
and teach .gitignore code how to use it.
add_excludes_from_file: clarify the bom skipping logic
Even though the previous step shifts where the "entry" begins, we
still iterate over the original buf[], which may begin with the
UTF-8 BOM we are supposed to be skipping. At the end of the first
line, the code grabs the contents of it starting at "entry", so
there is nothing wrong per-se, but the logic looks really confused.
Instead, move the buf pointer and shrink its size, to truly
pretend that UTF-8 BOM did not exist in the input.
dir: allow a BOM at the beginning of exclude files
Some text editors like Notepad or LibreOffice write an UTF-8 BOM in
order to indicate that the file is Unicode text rather than whatever the
current locale would indicate.
If someone uses such an editor to edit a gitignore file, we are left
with those three bytes at the beginning of the file. If we do not skip
them, we will attempt to match a filename with the BOM as prefix, which
won't match the files the user is expecting.
Signed-off-by: Carlos Martín Nieto <cmn@elego.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Merge branch 'jk/read-packed-refs-without-path-max' into maint
* jk/read-packed-refs-without-path-max:
read_packed_refs: use skip_prefix instead of static array
read_packed_refs: pass strbuf to parse_ref_line
read_packed_refs: use a strbuf for reading lines
is_hfs_dotgit: loosen over-eager match of \u{..47}
Our is_hfs_dotgit function relies on the hackily-implemented
next_hfs_char to give us the next character that an HFS+
filename comparison would look at. It's hacky because it
doesn't implement the full case-folding table of HFS+; it
gives us just enough to see if the path matches ".git".
At the end of next_hfs_char, we use tolower() to convert our
32-bit code point to lowercase. Our tolower() implementation
only takes an 8-bit char, though; it throws away the upper
24 bits. This means we can't have any false negatives for
is_hfs_dotgit. We only care about matching 7-bit ASCII
characters in ".git", and we will correctly process 'G' or
'g'.
However, we _can_ have false positives. Because we throw
away the upper bits, code point \u{0147} (for example) will
look like 'G' and get downcased to 'g'. It's not known
whether a sequence of code points whose truncation ends up
as ".git" is meaningful in any language, but it does not
hurt to be more accurate here. We can just pass out the full
32-bit code point, and compare it manually to the upper and
lowercase characters we care about.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Merge branch 'jk/rebuild-perl-scripts-with-no-perl-seting-change' into maint
The build procedure did not bother fixing perl and python scripts
when NO_PERL and NO_PYTHON build-time configuration changed.
* jk/rebuild-perl-scripts-with-no-perl-seting-change:
Makefile: have python scripts depend on NO_PYTHON setting
Makefile: simplify by using SCRIPT_{PERL,SH}_GEN macros
Makefile: have perl scripts depend on NO_PERL setting
"git checkout $treeish $path", when $path in the index and the
working tree already matched what is in $treeish at the $path,
still overwrote the $path unnecessarily.
* jk/checkout-from-tree:
checkout $tree: do not throw away unchanged index entries
Now that the index can block pathnames that can be mistaken
to mean ".git" on NTFS and FAT32, it would be helpful for
fsck to notice such problematic paths. This lets servers
which use receive.fsckObjects block them before the damage
spreads.
Note that the fsck check is always on, even for systems
without core.protectNTFS set. This is technically more
restrictive than we need to be, as a set of users on ext4
could happily use these odd filenames without caring about
NTFS.
However, on balance, it's helpful for all servers to block
these (because the paths can be used for mischief, and
servers which bother to fsck would want to stop the spread
whether they are on NTFS themselves or not), and hardly
anybody will be affected (because the blocked names are
variants of .git or git~1, meaning mischief is almost
certainly what the tree author had in mind).
Ideally these would be controlled by a separate
"fsck.protectNTFS" flag. However, it would be much nicer to
be able to enable/disable _any_ fsck flag individually, and
any scheme we choose should match such a system. Given the
likelihood of anybody using such a path in practice, it is
not unreasonable to wait until such a system materializes.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The point of disallowing ".git" in the index is that we
would never want to accidentally overwrite files in the
repository directory. But this means we need to respect the
filesystem's idea of when two paths are equal. The prior
commit added a helper to make such a comparison for NTFS
and FAT32; let's use it in verify_path().
We make this check optional for two reasons:
1. It restricts the set of allowable filenames, which is
unnecessary for people who are not on NTFS nor FAT32.
In practice this probably doesn't matter, though, as
the restricted names are rather obscure and almost
certainly would never come up in practice.
2. It has a minor performance penalty for every path we
insert into the index.
This patch ties the check to the core.protectNTFS config
option. Though this is expected to be most useful on Windows,
we allow it to be set everywhere, as NTFS may be mounted on
other platforms. The variable does default to on for Windows,
though.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
We do not allow paths with a ".git" component to be added to
the index, as that would mean repository contents could
overwrite our repository files. However, asking "is this
path the same as .git" is not as simple as strcmp() on some
filesystems.
On NTFS (and FAT32), there exist so-called "short names" for
backwards-compatibility: 8.3 compliant names that refer to the same files
as their long names. As ".git" is not an 8.3 compliant name, a short name
is generated automatically, typically "git~1".
Depending on the Windows version, any combination of trailing spaces and
periods are ignored, too, so that both "git~1." and ".git." still refer
to the Git directory. The reason is that 8.3 stores file names shorter
than 8 characters with trailing spaces. So literally, it does not matter
for the short name whether it is padded with spaces or whether it is
shorter than 8 characters, it is considered to be the exact same.
The period is the separator between file name and file extension, and
again, an empty extension consists just of spaces in 8.3 format. So
technically, we would need only take care of the equivalent of this
regex:
(\.git {0,4}|git~1 {0,3})\. {0,3}
However, there are indications that at least some Windows versions might
be more lenient and accept arbitrary combinations of trailing spaces and
periods and strip them out. So we're playing it real safe here. Besides,
there can be little doubt about the intention behind using file names
matching even the more lenient pattern specified above, therefore we
should be fine with disallowing such patterns.
Extra care is taken to catch names such as '.\\.git\\booh' because the
backslash is marked as a directory separator only on Windows, and we want
to use this new helper function also in fsck on other platforms.
A big thank you goes to Ed Thomson and an unnamed Microsoft engineer for
the detailed analysis performed to come up with the corresponding fixes
for libgit2.
This commit adds a function to detect whether a given file name can refer
to the Git directory by mistake.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now that the index can block pathnames that case-fold to
".git" on HFS+, it would be helpful for fsck to notice such
problematic paths. This lets servers which use
receive.fsckObjects block them before the damage spreads.
Note that the fsck check is always on, even for systems
without core.protectHFS set. This is technically more
restrictive than we need to be, as a set of users on ext4
could happily use these odd filenames without caring about
HFS+.
However, on balance, it's helpful for all servers to block
these (because the paths can be used for mischief, and
servers which bother to fsck would want to stop the spread
whether they are on HFS+ themselves or not), and hardly
anybody will be affected (because the blocked names are
variants of .git with invisible Unicode code-points mixed
in, meaning mischief is almost certainly what the tree
author had in mind).
Ideally these would be controlled by a separate
"fsck.protectHFS" flag. However, it would be much nicer to
be able to enable/disable _any_ fsck flag individually, and
any scheme we choose should match such a system. Given the
likelihood of anybody using such a path in practice, it is
not unreasonable to wait until such a system materializes.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The point of disallowing ".git" in the index is that we
would never want to accidentally overwrite files in the
repository directory. But this means we need to respect the
filesystem's idea of when two paths are equal. The prior
commit added a helper to make such a comparison for HFS+;
let's use it in verify_path.
We make this check optional for two reasons:
1. It restricts the set of allowable filenames, which is
unnecessary for people who are not on HFS+. In practice
this probably doesn't matter, though, as the restricted
names are rather obscure and almost certainly would
never come up in practice.
2. It has a minor performance penalty for every path we
insert into the index.
This patch ties the check to the core.protectHFS config
option. Though this is expected to be most useful on OS X,
we allow it to be set everywhere, as HFS+ may be mounted on
other platforms. The variable does default to on for OS X,
though.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
We do not allow paths with a ".git" component to be added to
the index, as that would mean repository contents could
overwrite our repository files. However, asking "is this
path the same as .git" is not as simple as strcmp() on some
filesystems.
HFS+'s case-folding does more than just fold uppercase into
lowercase (which we already handle with strcasecmp). It may
also skip past certain "ignored" Unicode code points, so
that (for example) ".gi\u200ct" is mapped ot ".git".
The full list of folds can be found in the tables at:
Implementing a full "is this path the same as that path"
comparison would require us importing the whole set of
tables. However, what we want to do is much simpler: we
only care about checking ".git". We know that 'G' is the
only thing that folds to 'g', and so on, so we really only
need to deal with the set of ignored code points, which is
much smaller.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
We complain about ".git" in a tree because it cannot be
loaded into the index or checked out. Since we now also
reject ".GIT" case-insensitively, fsck should notice the
same, so that errors do not propagate.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
We do not allow ".git" to enter into the index as a path
component, because checking out the result to the working
tree may causes confusion for subsequent git commands.
However, on case-insensitive file systems, ".Git" or ".GIT"
is the same. We should catch and prevent those, too.
Note that technically we could allow this for repos on
case-sensitive filesystems. But there's not much point. It's
unlikely that anybody cares, and it creates a repository
that is unexpectedly non-portable to other systems.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
read-tree: add tests for confusing paths like ".." and ".git"
We should prevent nonsense paths from entering the index in
the first place, as they can cause confusing results if they
are ever checked out into the working tree. We already do
so, but we never tested it.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
unpack-trees: propagate errors adding entries to the index
When unpack_trees tries to write an entry to the index,
add_index_entry may report an error to stderr, but we ignore
its return value. This leads to us returning a successful
exit code for an operation that partially failed. Let's make
sure to propagate this code.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
read_packed_refs: use skip_prefix instead of static array
We want to recognize the packed-refs header and skip to the
"traits" part of the line. We currently do it by feeding
sizeof() a static const array to strncmp. However, it's a
bit simpler to just skip_prefix, which expresses the
intention more directly, and without remembering to account
for the NUL-terminator in each sizeof() call.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Current code uses a fixed PATH_MAX-sized buffer for reading
packed-refs lines. This is a reasonable guess, in the sense
that git generally cannot work with refs larger than
PATH_MAX. However, there are a few cases where it is not
great:
1. Some systems may have a low value of PATH_MAX, but can
actually handle larger paths in practice. Fixing this
code path probably isn't enough to make them work
completely with long refs, but it is a step in the
right direction.
2. We use fgets, which will happily give us half a line on
the first read, and then the rest of the line on the
second. This is probably OK in practice, because our
refline parser is careful enough to look for the
trailing newline on the first line. The second line may
look like a peeled line to us, but since "^" is illegal
in refnames, it is not likely to come up.
Still, it does not hurt to be more careful.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
for_each_reflog_ent_reverse: turn leftover check into assertion
Our loop should always process all lines, even if we hit the
beginning of the file. We have a conditional after the loop
ends to double-check that there is nothing left and to
process it. But this should never happen, and is a sign of a
logic bug in the loop. Let's turn it into a BUG assertion.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
for_each_reflog_ent_reverse: fix newlines on block boundaries
When we read a reflog file in reverse, we read whole chunks
of BUFSIZ bytes, then loop over the buffer, parsing any
lines we find. We find the beginning of each line by looking
for the newline from the previous line. If we don't find
one, we know that we are either at the beginning of
the file, or that we have to read another block.
In the latter case, we stuff away what we have into a
strbuf, read another block, and continue our parse. But we
missed one case here. If we did find a newline, and it is at
the beginning of the block, we must also stuff that newline
into the strbuf, as it belongs to the block we are about to
read.
The minimal fix here would be to add this special case to
the conditional that checks whether we found a newline.
But we can make the flow a little clearer by rearranging a
bit: we first handle lines that we are going to show, and
then at the end of each loop, stuff away any leftovers if
necessary. That lets us fold this special-case in with the
more common "we ended in the middle of a line" case.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
push: truly use "simple" as default, not "upstream"
The plan for the push.default transition had all along been
to use the "simple" method rather than "upstream" as a
default if the user did not specify their own push.default
value. Commit 11037ee (push: switch default from "matching"
to "simple", 2013-01-04) tried to implement that by moving
PUSH_DEFAULT_UNSPECIFIED in our switch statement to
fall-through to the PUSH_DEFAULT_SIMPLE case.
When the commit that became 11037ee was originally written,
that would have been enough. We would fall through to
calling setup_push_upstream() with the "simple" parameter
set to 1. However, it was delayed for a while until we were
ready to make the transition in Git 2.0.
And in the meantime, commit ed2b182 (push: change `simple`
to accommodate triangular workflows, 2013-06-19) threw a
monkey wrench into the works. That commit drops the "simple"
parameter to setup_push_upstream, and instead checks whether
the global "push_default" is PUSH_DEFAULT_SIMPLE. This is
right when the user has explicitly configured push.default
to simple, but wrong when we are a fall-through for the
"unspecified" case.
We never noticed because our push.default tests do not cover
the case of the variable being totally unset; they only
check the "simple" behavior itself.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
sha1_name: avoid unnecessary sha1 lookup in find_unique_abbrev
An example where this happens is when doing an ls-tree on a tree that
contains a commit link. In that case, find_unique_abbrev is called
to get a non-abbreviated hex sha1, but still, a lookup is done as
to whether the sha1 is in the repository (which ends up looking for
a loose object in .git/objects), while the result of that lookup is
not used when returning a non-abbreviated hex sha1.
Signed-off-by: Mike Hommey <mh@glandium.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git add foo bar" adds neither foo nor bar when bar is ignored, but dies
to let the user recheck their command invocation. This becomes less
helpful when "git add foo.*" is subject to shell expansion and some of
the expanded files are ignored.
"git add --ignore-errors" is supposed to ignore errors when indexing
some files and adds the others. It does ignore errors from actual
indexing attempts, but does not ignore the error "file is ignored" as
outlined above. This is unexpected.
Change "git add foo bar" to add foo when bar is ignored, but issue
a warning and return a failure code as before the change.
That is, in the case of trying to add ignored files we now act the same
way (with or without "--ignore-errors") in which we act for more
severe indexing errors when "--ignore-errors" is specified.
Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net> Reviewed-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
If the user specifiers "normal" for a foreground color, this
should be a noop (while this may sound useless, it is the
only way to specify an unchanged foreground color followed
by a specific background color).
We also check that color "-1" does the same thing. This is
not documented, but has worked forever, so let's make sure
we keep supporting it.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
config: fix parsing of "git config --get-color some.key -1"
Most of git-config's command line options use OPT_BIT to
choose an action, and then parse the non-option arguments
in a context-dependent way. However, --get-color and
--get-colorbool are unlike the rest of the options, in that
they are OPT_STRING, taking the option name as a parameter.
This generally works, because we then use the presence of
those strings to set an action bit anyway. But it does mean
that the option-parser will continue looking for options
even after the key (because it is not a non-option; it is an
argument to an option). And running:
git config --get-color some.key -1
(to use "-1" as the default color spec) will barf, claiming
that "-1" is not an option. Instead, we should treat
--get-color and --get-colorbool as action bits, just like
--add, --get, and all the other actions, and then check that
the non-option arguments we got are sane. This fixes the
weirdness above, and makes those two options like all the
others.
This "fixes" a test in t4026, which checked that feeding
"-2" as a color should fail (it does fail, but prior to this
patch, because parseopt barfed, not because we actually ever
tried to parse the color).
This also catches other errors, like:
git config --get-color some.key black blue
which previously silently ignored "blue" (and now will
complain that you gave too many arguments).
There are some possible regressions, though. We now disallow
these, which currently do what you would expect:
# specifying other options after the action
git config --get-color some.key --file whatever
# using long-arg syntax
git config --get-color=some.key
However, we have never advertised these in the
documentation, and in fact they did not work in some older
versions of git. The behavior was apparently switched as an
accidental side effect of d64ec16 (git config: reorganize to
use parseopt, 2009-02-21).
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Ralf reported that '--recurse-submodules' option in push.c should not be
translated [1]. Before his commit is merged, remove superfluous
translations for push.c.
In order to catch up with the release of Git 2.2.0 final, make a batch
l10n update for the new l10n change brought by commit d52adf1 (trailer:
display a trailer without its trailing newline).
gitweb: hack around CGI's list-context param() handling
As of CGI.pm's 4.08 release, the behavior to call
CGI::param() in a list context is deprecated (because it can
be potentially unsafe if called inside a hash constructor).
This causes gitweb to issue a warning for some of our code,
which in turn causes the tests to fail.
Our use is in fact _not_ one of the dangerous cases, as we
are intentionally using a list context. The recommended
route by 4.08 is to use the new CGI::multi_param() call to
make it explicit that we know what we are doing.
However, that function is only available in 4.08, which is
about a month old; we cannot rely on having it.
One option would be to set $CGI::LIST_CONTEXT_WARN globally,
which turns off the warning. However, that would eliminate
the protection these newer releases are trying to provide.
We want to annotate each site as OK using the new function.
So instead, let's check whether CGI provides the
multi_param() function, and if not, provide an
implementation that just wraps param(). That will work on
both old and new versions of CGI. Sadly, we cannot just
check defined(\&CGI::multi_param), because CGI uses the
autoload feature, which claims that all functions are
defined. Instead, we just do a version check.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Makefile: have python scripts depend on NO_PYTHON setting
Like the perl scripts, python scripts need a dependency to ensure they
are rebuilt when switching between the "dummy" versions that run
without Python and the real thing.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Makefile: simplify by using SCRIPT_{PERL,SH}_GEN macros
SCRIPT_PERL_GEN is defined as $(patsubst %.perl,%,$(SCRIPT_PERL))
for use in targets like build-perl-script used by makefiles in
subdirectories that override SCRIPT_PERL (see v1.8.2-rc0~17^2,
"git-remote-mediawiki: use toplevel's Makefile", 2013-02-08).
The same expression is used in the rules that actually write the
generated perl scripts, and since these rules were introduced before
SCRIPT_PERL_GEN, they use the longhand instead of that macro. Use the
macro to make reading easier.
Likewise for SCRIPT_SH_GEN. The Python rules already got the same
simplification in v1.8.4-rc0~162^2~8 (2013-05-24).
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Reviewed-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Git-cvsimport is written in perl, which understandably
causes the tests to fail if you build with NO_PERL (which
will avoid building cvsimport at all). The earlier cvsimport
tests in t9600-t9602 are all marked with a PERL
prerequisite, but these ones are not.
The one in t9603 was likely not noticed because it is an
expected failure anyway.
The ones in t9604 have been around for a long time, but it
is likely that the combination of NO_PERL and having cvsps
installed is rare enough that nobody noticed.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
t0090: mark add-interactive test with PERL prerequisite
The add-interactive system is built in perl. If you build
with NO_PERL, running "git commit --interactive" will exit
with an error and the test will fail.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Makefile: have perl scripts depend on NO_PERL setting
If NO_PERL is not set, our perl scripts are built as
usual. If it is set, then we build "dummy" versions that
tell you git was built without perl support and exit
gracefully.
However, if you switch to NO_PERL in a directory with
existing build artifacts, we do not notice that the files
need rebuilt. We see only that they are newer than the
"unimplemented.sh" wrapper and assume they are done. So
doing:
make
make NO_PERL=Nope
would result in a git-add--interactive script that uses perl
(and running the test suite would make use of it).
Instead, we should trigger a rebuild of the perl scripts
anytime NO_PERL changes.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
create_default_files(): don't set u+x bit on $GIT_DIR/config
Since time immemorial, the test of whether to set "core.filemode"
has been done by trying to toggle the u+x bit on $GIT_DIR/config,
which we know always exists, and then testing whether the change
"took". I find it somewhat odd to use the config file for this
test, but whatever.
The test code didn't set the u+x bit back to its original state
itself, instead relying on the subsequent call to git_config_set()
to re-write the config file with correct permissions.
But ever since
daa22c6f8d config: preserve config file permissions on edits (2014-05-06)
git_config_set() copies the permissions from the old config file to
the new one. This is a good change in and of itself, but it
invalidates the create_default_files()'s assumption, causing "git
init" to leave the executable bit set on $GIT_DIR/config.
Reset the permissions on $GIT_DIR/config when we are done with the
test in create_default_files().
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Merge branch 'master' of git://github.com/git-l10n/git-po
* 'master' of git://github.com/git-l10n/git-po:
l10n: de.po: translate 62 new messages
l10n: de.po: Fixup one translation
l10n: de.po: use imperative form for command options
The git-push manual page used "gitlink" in one place instead of
"linkgit". Fix this so the link renders correctly.
Noticed-by: Dan Allen <dan.j.allen@gmail.com> Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Windows: correct detection of EISDIR in mingw_open()
According to the Linux open(2) man page, open() must return EISDIR
if a directory was attempted to be opened for writing. Our emulation
in mingw_open() does not get this right: it checks only for O_CREAT.
Fix it to check for a write request.
This fixes a failure in reflog handling, which opens files with
O_APPEND|O_WRONLY, but without O_CREAT, and expects EISDIR when the
named file happens to be a directory.
Signed-off-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>