Httpclient internally stores a copy of the certificate_check callback and
payload. When connecting via HTTPS, and if the server sends back
"Connection: close" after the first request, the following request would
attempt to re-use the httpclient and call the (now outdated) callback. In
particular for pygit2 this is a problem, since callbacks / payloads are only
valid for the duration of a libgit2 call, leading to a ffi.from_handle()
error and crashing the Python interpreter.

The path to the default CLI output has changed, update it.

Support non-cmake builds with an in-tree `experimental.h`

Not everybody builds libgit2 using cmake; provide an `experimental.h`
with no experiments configured for those that do not. To support this,
we also now create compile definitions for experimental functionality,
to supplant that empty `experimental.h`. cmake will continue to generate
the proper `experimental.h` file for use with `make install`.

#6366: When a worktree is missing, return GIT_ENOTFOUND.

Actually `cl_skip` the sha256 tests when we're not compiled for sha256,
instead of passing them.

Don't fail the whole clone if you can't find a default branch

Delete create.c.bak

Fix memory leak

Fixes #6365 : Uppercase windows.h include fails build in case-sensitive OS

fix compile on Windows with -DWIN32_LEAN_AND_MEAN

winhttp: support long custom headers

Test that we can successfully clone a repository that is namespace
scoped on the remote and does not advertise a HEAD. To do this, we must
specify the branch to checkout.

Test that we can successfully clone a repository that is namespace
scoped to a bare repository locally. We need not specify a checkout
branch in this case (obviously, since we do not check anything out in a
bare clone).

Only allow the remote default branch checking to fail when the remote
default branch doesn't exist (`git_remote__default_branch` returns
`GIT_ENOTFOUND`). If there was any other type of error - for example, an
allocation failure - we should not swallow that and continue to fail.

This allows us to support the case when a remote has not advertised a
HEAD -- this is possible, for example, when the remote has constrained
the caller to a particular namespace. But other remote failures remain
as failures.

CI Improvements

We previously (correctly) cleaned up the git daemon and SSH server, but
failed to clean up our bespoke HTTP server and HTTP proxies. Capture
their PIDs on process creation and kill them when we shut down.

GitHub has deprecated macOS 10.15; move to their new macOS 11 build
servers.

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

ensure the needed wincrypt.h is included

In commit 6bb35878 ("clone: set refs/remotes/origin/HEAD to default
branch when branch is specified, attempt 2") libgit2 was changed to set
the default remote branch when one was copied.

But it makes update_head_to_branch() return an error if the origin
doesn't even *have* a HEAD in the first place, since
git_remote_default_branch() will fail.

That's entirely wrong, and means that you cannot do "git_clone()" of a
particular branch on a remote repository when that remote doesn't have a
default branch at all.

So don't set the error code.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

clar: remove ftrunacte from libgit2 tests

tests: skip flaky-ass googlesource tests

sha256: indirection for experimental functions

The experimental function signature is only available when
`GIT_EXPERIMENTAL_SHA256` is enabled.

RFC: SHA256 proof of concept

meta: update version number to v1.6.0-alpha

v1.5.0