~vdupras/duskos

This repository has moved
arm: allow +) to be larger than 12b offsets
usermode: fix makefiles for packages

The "packages" targets of the makefile were outdated.
doc/cc: move to doc/comp/c

To be consistent with the rest of the doc hierarchy.
usermode arm: marking as working!

All tests pass on all 3 flavors on a rpi3 w/ NetBSD earmv7hf. They also run fine
on "stream" on a rpi1 w/ NetBSD earmv6hf.

Michael's new gradient test works fine on dusk-sdl.
usermode: fix SDL2 linking of dusk-sdl under BSDs

Without this fix, dusk-sdl compiles fine but can only be ran with
LD_LIBRARY_PATH=/usr/pkg/lib.
arm: guard "cidx" against null ranges

Also, better document, in doc/usage/lit, that "range words" usually handle null
ranges gracefully.

All tests in Usermode Dusk ARM now pass!
usermode arm: align RSP to 16b in syscallA

This makes dusk-curses work.
usermode arm: almost passing tests!

I'm getting really far with "make testum" under arm.

Still no dusk-curses and dusk-sdl though.
usermode arm: We have prompt!

Christmas miracles *do* exist! I was bracing for complicated I-cache
debugging but it turned out that what blocked me earlier was a simple
absolute reference I had missed in xcomp/arm/kernel for /mod
compilation.

With this fix, I get to prompt on dusk-stream!
xcomp/boot: remove clever ";" trick

I was happy about my cleverness, but this isn't I-cache friendly and
I'd like to have ARM working...
comp/c: fix typecast miscompilation
text/ed: fix multiline pasting bug

Previously, pasting contents with a LF in the middle of another line would
insert the first line in the middle the target line, then the other lines under
it.

Now, we have the expected behavior, that is, to remove the right part of the
target line, paste the whole contents, then re-add the previously removed part
at the end of the pasted contents.
sys/scratch: move to lib/scratch

This subsystem was created when the definition of "subsystem" was still fuzzy.
Now it's clearer and sys/scratch is not a subsystem.
usermode: fix broken include in my new "execchk"
Usermode SDL: Fix missing graphics updates

Seems that the event queue has a fixed size and when it is full, events
are dropped (SDL_PushEvent returns an error). Therefore, when
SDL_PushEvent returns an error wait until the event queue is (or has
been) empty, then try again, and if it errors again, show an error
message.

This fixes missing pixels/rows caused by Forth code that draws many
pixels. Added a "manual" test case that draws 2D gradients in different
drawing orders - no pixels should be left out.
usermode: add "execchk" program

It's an easier way to diagnose the capabilities of an OS to supply a RWX memory
area than running the whole Dusk.

Also, replace the "mflags* files with make includes. It's more flexible.
posix: fix spurious "Broken stack!" message in "dbg"

Empty stack is not a broken one...
usermode arm: third step!

We can see that INPTR and PSP work fine by spitting the first 42 input
characters through the wrapper and then exit cleanly.
usermode i386: fix ESP alignment during syscallA

Thanks Michael. I took the liberty to use another register to save ESP to save
a push of EBP, I think I'm in the clear.
Next