| Commit message (Collapse) | Author | Age | Files | Lines |
... | |
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
| |
As we add high-level constructs we'll start labeling low-level
constructs as unsafe, and highlighting them in red in our editor.
|
|
|
|
|
|
|
| |
No need for it once the program's loaded. And we keep programs from
running the header as code.
This also simplifies the header computation in the translator.
|
|
|
|
| |
Clarify a few happy accidents.
|
|
|
|
|
| |
The source 'language' is still entirely open. We'll see how it evolves
as I write programs in machine code.
|
|
|
|
| |
Give subx too the recent support for running a single test.
|
|
|
|
| |
Fix CI. Looks like 'std::' sometimes doesn't work.
|
|
|
|
| |
Make prints uniform.
|
| |
|
|
|
|
| |
Hopefully I won't need much more than exit, read and write.
|
| |
|
|
|
|
|
| |
We're now parsing the ELF spec more closely and better handling multiple
program header table entries.
|
| |
|
|
|
|
|
| |
I don't know how to tell nasm to generate an imm32. It's a minor stepping-stone
anyway; just emit the machine code directly.
|
|
|
|
| |
Turns out it was an open question I never got around to answering.
|
|
|
|
| |
Here's a few test binaries generated on 32-bit Linux.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
New plan: spend some time learning to program in machine code atop subx,
relying solely on a tiny subset of kernel-provided syscalls. Gradually
introduce helpers.
Helpers we're sure we don't need, so far:
a) Nested expressions
b) Garbage collection
c) One-size-fits-all memory allocation primitive
d) Function overloading and generics
Helpers we're sure we need, so far:
a) Dependency-injected versions of syscalls
b) Tangling directives
c) Statically checked types
Workflow for a C translator from ascii to binary:
a) run generated machine code atop subx (unit tests probably go here)
b) emit machine code packaged as an ELF file
c) check that the ELF binary runs natively
d) check that the ELF binary can be unwrapped and run atop subx
This is different from a conventional compiler because the 'HLL' is
unconstrained.
It is also different from Forth given the emphasis on types.
We want a simple stack that also encourages code sharing between
programmers. Conventional languages grow monotonically complex. Forth
discourages code sharing; it is non-trivial to figure out the 'shape' of
data a strange function expects on the stack.
|
|
|
|
| |
Fix CI.
|
|
|
|
| |
Fix CI.
|
| |
|
|
|
|
|
|
| |
So far we only checked if a single recipe used a variable with multiple
types in any single space. Now we also ensure that the types deduced for
a variable in a space are identical across recipes.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
| |
Seems incredible that this is all it took. Needs more testing.
I also need to rethink how we organize our layers about addresses.
Alloc-id stuff is scattered everywhere. The space for alloc-ids is
perhaps unavoidably scattered. Just assume the layout from the start.
But it seems bad that the scenario testing the lookup-time validation is
in the 'abandon' layer when the code is in the 'lookup' layer.
|
|
|
|
| |
Tautological for now since all alloc-ids are zero.
|
|
|
|
| |
Drop a stray hunk that is obsoleted by 'deaddress'.
|
|
|
|
| |
This has taken me almost 6 weeks :(
|
|
|
|
| |
Standardize use of type ingredients some more.
|
|
|
|
| |
Undo the relayout of 4259.
|
|
|
|
|
| |
Implement literal constants before type abbreviations, reducing some
unnecessary tangling.
|
| |
|
|
|
|
|
|
|
|
|
| |
They uncovered one bug: in edit/003-shortcuts.mu
<scroll-down> was returning 0 for an address in one place where I
thought it was returning 0 for a boolean.
Now we've eliminated this bad interaction between tangling and punning
literals.
|
|
|
|
|
| |
'deaddress' is a terrible name. Hopefully I'll come up with something
better.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
I've been working on this slowly over several weeks, but it's too hard
to support 0 as the null value for addresses. I constantly have to add
exceptions for scalar value corresponding to an address type (now
occupying 2 locations). The final straw is the test for 'reload':
x:num <- reload text
'reload' returns an address. But there's no way to know that for
arbitrary instructions.
New plan: let's put this off for a bit and first create support for
literals. Then use 'null' instead of '0' for addresses everywhere. Then
it'll be easy to just change what 'null' means.
|
|
|
|
|
| |
We have some ugly duplication in computing size_of on containers between
layers 30/33 and 55.
|
| |
|
| |
|
|
|
|
| |
We've had this ability for Mu scenarios forever.
|
| |
|
|
|
|
| |
Also allow running a single test, to speed things up still further.
|
|
|
|
| |
Avoid modifying memory *before* the null check.
|
|
|
|
| |
Why do we have this silent null check? All tests pass without it.
|