Currently, the reflog disallows any entries that have a message with
newlines, as that would effectively break the reflog format, which may
contain a single line per entry, only. Upstream git behaves a bit
differently, though, especially when considering stashes: instead of
rejecting any reflog entry with newlines, git will simply replace
newlines with spaces. E.g. executing 'git stash push -m "foo\nbar"' will
create a reflog entry with "foo bar" as entry message.

This commit adjusts our own logic to stop rejecting commit messages with
newlines. Previously, this logic was part of `git_reflog_append`, only.
There is a second place though where we add reflog entries, which is the
serialization code in the filesystem refdb. As it didn't contain any
sanity checks whatsoever, the refdb would have been perfectly happy to
write malformatted reflog entries to the disk. This is being fixed with
the same logic as for the reflog itself.

The code style of `git_stash_save` doesn't really match our current
coding style. Update it to match our current policies more closely.

Fix file locking on POSIX OS

cmake: update minimum CMake version to v3.5.1

The flag GIT_FILEBUF_FORCE currently does two things:
     1. It will cause the filebuf to create non-existing leading
        directories for the file that is about to be written.
     2. It will forcibly remove any pre-existing locks.
While most call sites actually do want (1), they do not want to
remove pre-existing locks, as that renders the locking mechanisms
effectively useless.
Introduce a new flag `GIT_FILEBUF_CREATE_LEADING_DIRS` to
separate both behaviours cleanly from each other and convert
callers to use it instead of `GIT_FILEBUF_FORCE` to have them
honor locked files correctly.

As this conversion removes all current users of `GIT_FILEBUF_FORCE`,
this commit removes the flag altogether.

patch_parse: handle patches with new empty files

Back in commit cf9f3452 (cmake: bump minimum version to 2.8.11,
2017-09-06), we have bumped the minimum CMake version to require at
least v2.8.11. The main hold-backs back then were distributions like
RHEL/CentOS as well as Ubuntu Trusty, which caused us to not target a
more modern version. Nowadays, Ubuntu Trusty has been EOL'd and CentOS 6
has CMake v3.6.1 available via the EPEL6 repository, and thus it seems
fair to upgrade to a more recent version.

Going through repology [1], one can see that all supported mainstream
distributions do in fact have CMake 3 available. Going through the list,
the minimum version that is supported by all mainstream distros is in
fact v3.5.1:

	- CentOS 6 via EPEL6: 3.6.1
	- Debian Oldstable: 3.7.2
	- Fedora 26: 3.8.2
	- OpenMandriva 3.x: 3.5.1
	- Slackware 14.2: 3.5.2
	- Ubuntu 16.04: 3.5.1

Consequentally, let's upgrade CMake to the minimum version of 3.5.1 and
remove all the version CMake checks that aren't required anymore.

[1]: https://repology.org/project/cmake/versions

DRY commit parsing

The commit list's in- and out-degrees are currently stored as `unsigned
short`. When assigning it the value of `git_array_size`, which returns
an `size_t`, this generates a warning on some Win32 platforms due to
loosing precision.

We could just cast the returned value of `git_array_size`, which would
work fine for 99.99% of all cases as commits typically have less than
2^16 parents. For crafted commits though we might end up with a wrong
value, and thus we should definitely check whether the array size
actually fits into the field.

To ease the check, let's convert the fields to store the degrees as
`uint16_t`. We shouldn't rely on such unspecific types anyway, as it may
lead to different behaviour across platforms. Furthermore, this commit
introduces a new `git__is_uint16` function to check whether it actually
fits -- if not, we return an error.

This allows us to pick which data from a commit we're interested in.
This will be used by the revwalk code, which is only interested in
parents' and committer data.

azure: avoid building and testing in Docker as root

regexp: implement a new regular expression API

Patches containing additions of empty files will not contain diff data
but will end with the index header line followed by the terminating
sequence "-- ". We follow the same logic as in cc4c44a9 and allow "-- "
to immediately follow the index header.

git_refdb API fixes

Don't use enum for flags

Using an `enum` causes trouble when used with C++ as bitwise operations are not possible w/o casting (e.g., `opts.flags &= ~GIT_BLOB_FILTER_CHECK_FOR_BINARY;` is invalid as there is no `&=` operator for `enum`).

Signed-off-by: Sven Strickroth <email@cs-ware.de>

Right now, all tests in libgit2's CI are being executed as root
user. As libgit2 will usually not run as a root user in "normal"
usecases and furthermore as there are tests that rely on the
ability to _not_ be able to create certain paths, let's instead
create an unprivileged user "libgit2" and use that across all
docker images.

valgrind: suppress memory leaks in libssh2_session_handshake

On Ubuntu, the combination of libgcrypt and libssh2 is quite old and
known to contain memory leaks. We thus have several functions listed in
our suppressions file that are known to leak. Due to a recent update of
libssh2 or libgcrypt, there now are new memory leaks caused by
libssh2_session_handshake and libssh2_init that cause the CI to fail.

Add a new suppression to fix the issue.

buffer: fix writes into out-of-memory buffers

The old POSIX regex wrappers have been superseded by our own regexp API
that provides a higher-level abstraction. Remove the POSIX wrappers in
favor of the new one.

The old POSIX regex API has been superseded by our new regexp API.
Convert all users to make use of the new one.

We currently support a set of different regular expression backends with
PCRE, PCRE2, regcomp(3P) and regcomp_l(3). The current implementation of
this is done via a simple POSIX wrapper that either directly uses
supplied functions or that is a very small wrapper.

To support PCRE and PCRE2, we use their provided <pcreposix.h> and
<pcre2posix.h> wrappers. These wrappers are implemented in such a way
that the accompanying libraries pcre-posix and pcre2-posix provide the
same symbols as the libc ones, namely regcomp(3P) et al. This works out
on some systems just fine, most importantly on glibc-based ones, where
the regular expression functions are implemented as weak aliases and
thus get overridden by linking in the pcre{,2}-posix library. On other
systems we depend on the linking order of libc and pcre library, and as
libc always comes first we will end up with the functions of the libc
implementation. As a result, we may use the structures `regex_t` and
`regmatch_t` declared by <pcre{,2}posix.h>, but use functions defined by
the libc, leading to segfaults.

The issue is not easily solvable. Somed distributions like Debian have
resolved this by patching PCRE and PCRE2 to carry custom prefixes to all
the POSIX function wrappers. But this is not supported by upstream and
thus inherently unportable between distributions. We could instead try
to modify linking order, but this starts becoming fragile and will not
work e.g. when libgit2 is loaded via dlopen(3P) or similar ways. In the
end, this means that we simply cannot use the POSIX wrappers provided by
the PCRE libraries at all.

Thus, this commit introduces a new regular expression API. The new API
is on a tad higher level than the previous POSIX abstraction layer, as
it tries to abstract away any non-portable flags like e.g. REG_EXTENDED,
which has no equivalents in all of our supported backends. As there are
no users of POSIX regular expressions that do _not_ reguest REG_EXTENDED
this is fine to be abstracted away, though. Due to the API being
higher-level than before, it should generally be a tad easier to use
than the previous one.

Note: ideally, the new API would've been called `git_regex_foobar` with
a file "regex.h" and "regex.c". Unfortunately, this is currently
impossible to implement due to naming clashes between the then-existing
"regex.h" and <regex.h> provided by the libc. As we add the source
directory of libgit2 to the header search path, an include of <regex.h>
would always find our own "regex.h". Thus, we have to take the bitter
pill of adding one more character to all the functions to disambiguate
the includes.

To improve guarantees around cross-backend compatibility, this commit
also brings along an improved regular expression test suite
core::regexp.

Before printing into a `git_buf` structure, we always call `ENSURE_SIZE`
first. This macro will reallocate the buffer as-needed depending on
whether the current amount of allocated bytes is sufficient or not. If
`asize` is big enough, then it will just do nothing, otherwise it will
call out to `git_buf_try_grow`. But in fact, it is insufficient to only
check `asize`.

When we fail to allocate any more bytes e.g. via `git_buf_try_grow`,
then we set the buffer's pointer to `git_buf__oom`. Note that we touch
neither `asize` nor `size`. So if we just check `asize > targetsize`,
then we will happily let the caller of `ENSURE_SIZE` proceed with an
out-of-memory buffer. As a result, we will print all bytes into the
out-of-memory buffer instead, resulting in an out-of-bounds write.

Fix the issue by having `ENSURE_SIZE` verify that the buffer is not
marked as OOM. Add a test to verify that we're not writing into the OOM
buffer.

When growing buffers, we repeatedly multiply the currently allocated
number of bytes by 1.5 until it exceeds the requested number of bytes.
This has two major problems:

    1. If the current number of bytes is tiny and one wishes to resize
       to a comparatively huge number of bytes, then we may need to loop
       thousands of times.

    2. If resizing to a value close to `SIZE_MAX` (which would fail
       anyway), then we probably hit an infinite loop as multiplying the
       current amount of bytes will repeatedly result in integer
       overflows.

When reallocating buffers, one typically chooses values close to 1.5 to
enable re-use of resulting memory holes in later reallocations. But
because of this, it really only makes sense to use a factor of 1.5
_once_, but not looping until we finally are able to fit it. Thus, we
can completely avoid the loop and just opt for the much simpler
algorithm of multiplying with 1.5 once and, if the result doesn't fit,
just use the target size. This avoids both problems of looping
extensively and hitting overflows.

This commit also adds a test that would've previously resulted in an
infinite loop.

If growing a buffer fails, we set its pointer to the static
`git_buf__oom` structure. While we correctly free the old pointer if
`git__malloc` returned an error, we do not free it if there was an
integer overflow while calculating the new allocation size. Fix this
issue by freeing the pointer to plug the memory leak.

cred: add missing private header in GSSAPI block

Should have been part of 8bf0f7eb

CMake pkg-config modulification

Update chat resources in README.md

Circular header splitting

azure: build Docker images as part of the pipeline