Tests failing.
This approach seems wrong. I'm not sure even the tests are correct. Also,
some open questions:
1. Should setting the overflow flag always set the carry flag?
2. Should the carry flag only be set on add/subtract/compare, or by all
arithmetic ops?
3. Had to turn off the -ftrapv flag in `build`. Is there a way to detect
overflow without actually causing overflow?
Once we start setting CF correctly we have to implement jump above/below
instructions (8- and 32-bit displacement variants).
https://github.com/akkartik/mu/issues/30
I've been saying for a while[1][2][3] that adding extra abstractions makes
things harder for newcomers, and adding new notations doubly so. And then
I notice this DSL in my own backyard. Makes me feel like a hypocrite.
[1] https://news.ycombinator.com/item?id=13565743#13570092
[2] https://lobste.rs/s/to8wpr/configuration_files_are_canary_warning
[3] https://lobste.rs/s/mdmcdi/little_languages_by_jon_bentley_1986#c_3miuf2
The implementation of the DSL was also highly hacky:
a) It was happening in the tangle/ tool, but was utterly unrelated to tangling
layers.
b) There were several persnickety constraints on the different kinds of
lines and the specific order they were expected in. I kept finding bugs
where the translator would silently do the wrong thing. Or the error messages
sucked, and readers may be stuck looking at the generated code to figure
out what happened. Fixing error messages would require a lot more code,
which is one of my arguments against DSLs in the first place: they may
be easy to implement, but they're hard to design to go with the grain of
the underlying platform. They require lots of iteration. Is that effort
worth prioritizing in this project?
On the other hand, the DSL did make at least some readers' life easier,
the ones who weren't immediately put off by having to learn a strange syntax.
There were fewer quotes to parse, fewer backslash escapes.
Anyway, since there are also people who dislike having to put up with strange
syntaxes, we'll call that consideration a wash and tear this DSL out.
---
This commit was sheer drudgery. Hopefully it won't need to be redone with
a new DSL because I grow sick of backslashes.
I've extracted it into a separate binary, independent of my Mu prototype.
I also cleaned up my tracing layer to be a little nicer. Major improvements:
- Realized that incremental tracing really ought to be the default.
And to minimize printing traces to screen.
- Finally figured out how to combine layers and call stack frames in a
single dimension of depth. The answer: optimize for the experience of
`browse_trace`. Instructions occupy a range of depths based on their call
stack frame, and minor details of an instruction lie one level deeper
in each case.
Other than that, I spent some time adjusting levels everywhere to make
`browse_trace` useful.
Add the standard mnemonic for each opcode.
We aren't ever going to have complete docs of the subset of the x86 ISA
we support, so we need to help readers cross-correlate with the complete
docs.
One of the more painful things I had to debug with machine code. Tricks
I used can be seen in ex10.subx:
- printing argv[1] in various places
- printing a single 'X' in various places to count how many times we get
to different instructions
- exiting with the current value of EAX in various places
I repeatedly went down the wrong trail in several ways:
- forgetting that the problem lay in native runs, and accidentally switching
to subx runs during debugging.
- forgetting to pass commandline args, because ex10 doesn't check its argv
- writing the wrong comment for an instruction, and then miscalculating
the set of registers that need to be saved.
- forgetting that syscalls clobber EAX.
Debugging native runs is hard, because you have to write non-trivial code
to instrument the binary, and instrumentation can itself be buggy.
When we finally tracked it down, I recognized the problem immediately.
I'd meant to confirm the behavior of opcode 8a against bare metal, and
then forgot.
In any case, opcode 8a was inconsistent with 88. Sloppy.
Streamline the factorial function; we don't need to save a stack variable
into a register before operating on it. All instructions can take a stack
variable directly.
In the process we found two bugs:
a) Opcode f7 was not implemented correctly. It was internally consistent
but I'd never validated it against a natively running program. Turns out
it encodes multiple instructions, not just 'not'.
b) The way we look up imm32 operands was sometimes reading them before
disp8/disp32 operands.
The new example ex9 doesn't yet work natively.
In the process I've emulated the kernel's role in providing args, implemented
a couple of instructions acting on 8-bit operands (useful for ASCII string
operations), and begun the start of the standard library (ascii_length
is the same as strlen).
At the level of SubX we're just only going to support ASCII.
