When the stream writing function was written, it assume that
libssh2_channel_write() would always write all of the data to the
wire. This is only true for the first 32k of data, which it tries to
fit into one ssh packet.

Since it can perform short writes, call it in a loop like we do for
send(), advancing the buffer offset.

Don't use read-write accounts for the tests

netops: init OpenSSL once under lock

The assembla failure we were seeing referred to a private repository,
which is not what is there at the moment.

This reverts 1fd21b03

Don't write in plaintext the password of an account which has full
control over the repository. Instead use an account with read-only
access.

We want to check whether the credentials callback is NULL, not whether
the payload is.

Bring together all of the OpenSSL initialization to
git_threads_init() so it's together and doesn't need locks.

Moving it here also gives us libssh2 thread safety (when built against
openssl).

OpenSSL's tests init everything in the main thread, so let's do that.

When using in a multithreaded context, OpenSSL needs to lock, and leaves
it up to application to provide said locks.

We were not doing this, and it's just luck that's kept us from crashing
up to now.

Add docs for git_clone_local_t and move the docs for the
git_clone_options to each field.

The OpenSSL init functions are not reentrant, which means that running
multiple fetches in parallel can cause us to crash.

Use a mutex to init OpenSSL, and since we're adding this extra checks,
init it only once.

revwalk: more sensible array handling

Instead of using a sentinel empty value to detect the last commit, let's
check for when we get a NULL from popping the stack, which lets us know
when we're done.

The current code causes us to read uninitialized data, although only on
RHEL/CentOS 6 in release mode. This is a readability win overall.

treebuilder: insert sorted

By inserting in the right position, we can keep the vector sorted,
making entry insertion almost twice as fast.

Bump library version to 0.21.0 and SONAME to 21

Since the SOVERSION doesn't need to follow the library's version and
simply needs to be monotonically increasing whenever we release
something that breaks the ABI, we can set some number and allow multiple
versions of the library to be installed side-by-side.

We start here with the minor version as that's what we release for now,
and it allows to backport this change to earlier versions.

More remote rename fixes

Win32: Fix object::cache::threadmania test on x64

Win32 test fixes

This was a bad idea. Don't free except in the free function.

A symref inside the namespace gets renamed, we should make it point to
the target's new name.

This is for the origin/HEAD -> origin/master type of situations.

There is no reason why we need to use a callback here. A string array
fits better with the usage, as this is not an event and we don't need
anything from the user.

We must make sure that the name pointer remains valid, so make sure to
allocate the new one before freeing the old one and swap them so the
user never sees an invalid pointer.

We don't allow renames of anonymous remotes, so there's no need to
handle them.

A remote is always associated with a repository, so there's no need to
check for that.

Remote rename fixes

Update AUTHORS

Add me. :)

Tighten up which references we consider for renaming so we don't try to
rename unrelated ones and end up with unexplained references.

If there is a reference on the target namespace, git overwrites it, so
let's do the same.

When there is a reference in the target namespace, we should overwrite
it. Instead it gets a different name under the current code.