Add Linux's syscalls

[data-man]

I'll improve it later for more targets if this PoC is OK.

[lerno]

If I understand correctly it becomes like Syscall.READ.call(a, b, c), why shouldn't we simply have: linux::syscall_read(a, b, c)? linux::syscall_write(a, b c)

[data-man]

linux::syscall_read(a, b, c)? linux::syscall_write(a, b c)

Add hundreds of functions?

[data-man]

Or just add constdefs?

[lerno]

I don't see how using constants/enums would be better than macros?

[data-man]

Some std modules uses this pattern:

constdef Foo : inline uint
{
    FOO1        = 1,
    FOO2        = 2
}

const uint BAR1 = Foo.FOO1;
const uint BAR2 = Foo.FOO2;

So, what about this:

constdef SysCall : inline uint
{
    <* (unsigned int fd, char *buf, size_t count) *>
    READ         = 0,
    <* (unsigned int fd, const char *buf, size_t count) *>
    WRITE        = 1

    etc.
}

<* (unsigned int fd, char *buf, size_t count) *>
const uint SYS_READ = SysCall.READ;

<* (unsigned int fd, const char *buf, size_t count) *>
const uint SYS_WRITE = SysCall.WRITE;

etc.

[lerno]

That doesn't answer why it couldn't be macros? Wouldn't that be nicer?

[data-man]

That doesn't answer why it couldn't be macros? Wouldn't that be nicer?

How useful are hundreds of macros without ported Linux types to C3 types?

[lerno]

You are suggesting:

SYS_READ.call(fd, buf, count); 

With everything untyped.

I don't see why one shouldn't do

macro sys_read(uint, char* buf, usz count) { ... }

It seems like a gross misuse of macros and enums?

[lerno]

This PR has been dead now for over a month, what should we do with it?