When calling `git_submodule_update` on a submodule, we have to retrieve
the ID of the submodule entry in the index. If the function is called on
a submodule which is only partly initialized, the submodule entry may
not be added to the index yet. This leads to an assert when trying to
look up the blob later on.

Fix the issue by checking if the index actually holds the submodule's
ID and erroring out if it does not.

Skip uninteresting commits in revwalk timesort iterator

git_futils_readbuffer: don't compute sha-1

rebase: ignore untracked files in submodules

Fixes #4099

Don't compute the sha-1 in `git_futils_readbuffer_updated` unless the
checksum was requested.  This means that `git_futils_readbuffer` will
not calculate the checksum unnecessarily.

Support namespaced references again

These simple tests only ensure that we enforce the existence of a
namespace; these mirror the rugged tests, they are not exhaustive.

An untracked file in a submodule should not prevent a rebase from
starting.  Even if the submodule's SHA is changed, and that file would
conflict with a new tracked file, it's still OK to start the rebase
and discover the conflict later.

Signed-off-by: David Turner <dturner@twosigma.com>

git_commit_create: freshen tree objects in commit

Freshen the tree object that a commit points to during commit time.

Introduce (optional) SHA1 collision attack detection

We never set `SHA1_TYPE` to `builtin`.  Don't bother testing for it.

Include the SHA1 collision attack detection library from
https://github.com/cr-marcstevens/sha1collisiondetection

cmake: only enable supported compiler warning flags

We currently unconditionally enable the "-Wall" and "-Wextra" flags.
Some platforms rely on compilers which do not support these flags,
though. One of these platforms is Haiku, which does not support
"-Wextra" due to being stuck on GCC version 2.

Fix builds on such platforms by adding these flags only if supported by
the compiler.

tests: refs::create: fix memory leak

Fix: make reflog include "(merge)" for merge commits

This test ensures that the string '(merge)' is included in the reflog
when a merge commit is made.

This fixes issue #4094

tests: Add create__symbolic_with_arbitrary_content

Fix minor typos in CONVENTIONS.md

Fix inet_pton tests triggering an assert in Haiku

Haiku will assert in a nightly build if the "dst" input to inet_pton() is NULL.

Use C style comments as per style guide,
and fix mismatching indentation.

This test ensures that it's possible to create a symbolic ref that
has arbitrary data as its target. It also ensures it's possible
to obtain the target of that symbolic reference from the git_reference
object.

appveyor: don't rewrite the system mingw

Download mingw-w64 into our build directory and execute it there, don't
try to overwrite the system's mingw.

The 'appveyor' branch is useful for testing AppVeyor builds.

fix regression from #4092

Fixes a regression from #4092. This is a crash on 32-bit and I assume that
it doesn't do the right thing on 64-bit either. MSVC emits a warning for this,
but of course, it's easy to get lost among all of the similar 'possible loss
of data' warnings.

Attrcache cleanups

Provide error on gcc < 4.1 when THREADSAFE

Provide a descriptive error message when compiling THREADSAFE on gcc
versions < 4.1.  We require the atomic primitives (eg
`__sync_synchronize`) that were introduced in that version.

(Note, clang setes `__GNUC__` but appears to set its version > 4.1.)

Remove useless indirection from `git_attr_cache__init` to
`git_attr_cache__do_init`. The difference is that the
`git_attr_cache__init` macro first checks if the cache is already
initialized and, if so, not call `git_attr_cache__do_init`. But
actually, `git_attr_cache__do_init` already does the same thing and
returns immediately if the cache is already initialized.

Remove the indirection.