These headers have no associated C files, and are mainly used for
utility logic.
Having them in `/include` makes them more easily located, and more
globally used.
We are using "classic" header guards here over `#pragma once` for better
support. Since `#pragma once` is defined by the compiler, rather than
standard C.
Why isn't it used across the project? I have no idea.
Note: not changing `NBT_I*` macros, since the NBT specification does not
specify a difference between signed/unsigned types. Generally speaking,
internally we treat these as unsigned types for the purpose of
computations being correct.
- Improved logic for overriding compiler flags.
- Made tooling and logging more robust.
- Added functionality to disable "silent mode" by listening to
`--no-silent`
- Removed OSX/Darwin build support (for now)
- No longer utilising /obj directory, but instead just building files
right in the worktree.
- Generating dependency files seperately from general code, for
clairity.
The unit tests and continious integration is... exessive.
Yes, it's useful, but creates more complications than that it helps, so
it is removed for the time being. Maybe I will add a new framework at a
later date, but at the moment it is constraining development.
Mainly so the makefile is a bit less esoteric to use, debugging flags
must be specified manually.
changed:
- `$RM` for calling `rm`
- removed unused `$ISWIN`
- renamed `test` to `check`
- added `install` and `install-strip` recipes, which for now remain
empty
- added `-g` to `$CFLAGS`, since it helps debugging, and in case of
failures makes bug report less of a headache.
I decided to not modify it further so "specialised tools" don't need to
store files in the git repo, which seemed to complicate logic.
This is the same reason to why I chose to not adapt clean in a way to
introduce `mostlyclean`.
Functions now break before their brace, mirroring the Linux kernel.
The reason for this is that breaking the parameter list otherwise makes
code unreadable.
