"git_config_backend" have been renamed to "git_config_file", which
implements a generic interface to access a configuration file -- be it
either on disk, from a DB or whatever mumbojumbo.

I think this makes more sense.

We have a lot of utility methods that make path building trivial. Use
them!

A couple of config improvements

Fix a few minor typos in comments and error messages

Regarding "initialize" vs. "initialise", www.dict.cc says the first is American
English whereas the latter in British English. For consistency, we should
stick to American English.

index: Add API for unmerged entries

Move odb.c to the new error handling

test_lib: add return value to git_test

Consider environment-specified CFLAGS

Cmake msvc2010

Add missing free in git_odb_new().

Signed-off-by: schu <schu-github@schulog.org>

MSVC supports "inline" only in C++ code, not in C code.

MSVC has _str(n)icmp instead.

Save the return value of functions not passing must_pass() and report
the returned error.

Signed-off-by: schu <schu-github@schulog.org>

When I changed it over to use different strings for the variable and
the name, cvar_name_normalize was left behind. Fix this and rename to
cvar_normalize_name to reflect the incompatible change.

Signed-off-by: Carlos Martín Nieto <cmn@elego.de>

git__rethrow was missing an underscore.

Signed-off-by: Carlos Martín Nieto <cmn@elego.de>

tests: update NAMING file

Implement config parsing. Finally!

Take this opportunity to fix an instance of returning
GIT_EOBJCORRUPTED when malloc failed.

Signed-off-by: Carlos Martín Nieto <cmn@elego.de>

Now we use a simple closed-addressing cache. Cuckoo hashing was creating
too many issues with race conditions. Fuck that.

Let's see what happens performance wise, we may have to roll back or
come up with another way to implement an efficient multi-threaded cache.

Do not try to adquire the same node lock twice when the cuckoo hashing
resolves to the same node.