| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
| |
I'm sick of fighting vim's filetype detection. No modeline and files
highlight in random colors. I add a modeline and it stops highlighting
tangle comments. Even though it read my #$%# vimrc! Fuck this shite.
|
| |
|
|
|
|
|
| |
Why did I think STL's map wasn't efficient? It has logarithmic
complexity (maintains a tree internally) and is faster than hashing for
small containers. It's the more portable solution and should be what I
turn to by default.
|
| | |
|
| | |
|
| |
|
|
|
|
| |
Reorganized commandline handling to perform test space assertion in just
one place at the right layer. Hopefully this issue will not bite us
again.
|
| |
|
|
|
|
|
|
|
|
| |
Move assertions about test space right before running all tests. (But
not before running *every* test.)
This is a good trip-wire, but it requires that we remember to set it in
any new code paths running tests.
Still one wart: layer 1 depends on layer 10.
|
| |
|
|
|
|
|
| |
Required still more tweaking of Recipe namespaces. Mindlessly inserting
setup() took a couple of hours to debug because the test function and
the function it was testing ended up getting the same recipe number,
with the inevitable infinite loop :/
|
| |
|
|
|
|
| |
We have tangled vs regular comments, but they highlight the same. So
we're still at 3 colors for comments. Anything more starts to seem
gratuitous.
|
| | |
|
| |
|
|
|
| |
In the process I give up trace stability when I move files around, but I
gain in exchange the ability to move files around.
|
| |
|
|
|
|
|
|
|
| |
This seems more obvious to laypeople.
I've also come up with a design for sum types: I'm going to call them
exclusive containers. You call 'get' on containers, 'index' on arrays,
and 'maybe-convert' on exclusive containers (as well as tagged types,
but that's even later).
|
| | |
|
| | |
|
| |
|
|
| |
This was a pain to debug.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
| |
Things are quite intricate to avoid reloading all recipes before every
test. But mu wasn't really intended to sidestep intrinsic intricacy.
|
| | |
|
| |
|
|
| |
The layers are a mess now, but we'll take care of that later.
|
| | |
|
| |
|
|
|
| |
I think the string-equal scenarios are failing to fail. But we'll fix
them once we wrap up test isolation.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
I'm not testing for bounds-check errors yet, though.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
| |
I'm making two changes to how I compute field offsets:
a) I just replace offset names up front, before I even manage field
names. I don't bother disallowing x:integer and x:offset in the same
function. Let's see if that leads us astray. Certainly saves code.
b) I don't bother canonizing the first arg of a get since we know it
has to have a type that is some number of 'address' followed by a
record. Just assume that we have the right number of 'deref's.
|
| | |
|
| |
|
|
| |
Cleaner way to signal that a reagent's value has been computed.
|
| |
|
|
| |
Addresses for reagents are now computed after all transforms.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
| |
These #defines and references now span many different layers. Let's see
if the lack of encapsulation causes problems.
Also interesting to run into a case where I need to modify a
foundational layer and touch every single scenario/trace. Only
alternative was to duplicate all the different layers that add
instructions. Sign of problems with this model?
|
| | |
|
| | |
|