SrcFn represents the function in the linked output file, if the
`Decl` is a function. This is stored here and not in `Fn` because `Decl`
survives across updates but `Fn` does not.
TODO Look into making `Fn` a longer lived structure and moving this field there
to save on memory usage.
Empirically, debug info consumers do not respect this field, or otherwise
consider it to be an error when it does not point exactly to the end of the header.
Therefore we rely on the NOP jump at the beginning of the Line Number Program for
padding rather than this field.
llvm-dwarfdump says the line number data is fine; gdb and
binutils-readelf crap out.
Supports writing the first function. Still TODO is:
* handling the .debug_line header growing too large
* adding a new file to an existing compilation
* adding an additional function to an existing file
* handling incremental updates
* adding the main IR debug ops for IR instructions
There are also issues to work out:
* readelf --debug-dump=rawline is saying there is no .debug_str section
even though there is
* readelf --debug-dump=decodedline is saying the file index 0 is bad
and reporting some other kind of corruption.
* the .debug_line header is written properly
* link.File.Elf gains:
- SrcFn, which is now a field in Module.Fn
- SrcFile, which is now a field in Module.Scope.File
* link.File.Elf gets a whole *Package field rather than only
root_src_dir_path.
* the fields first_dbg_line_file and last_dbg_line_file tell where the
Line Number Program begins and ends, which alows moving files when
the header gets too big, and allows appending files to the end.
* codegen is passed a buffer for emitting .debug_line
Line Number Program opcodes for functions.
See #5963
There is some work-in-progress code here, but I need to go make some
experimental changes to changing how to represent source locations and I
want to do that in a separate commit.
* multiple returns jump to one canonical function exitlude. This is in
preparation for the defer feature.
* simple elision of trivial jump relocs.
* omit prelude/exitlude for naked calling convention functions.
* fix not switching on arch for prelude/exitlude
* fix swapped registers when setting stack mem from a register
* `optimize_mode` is passed to `link.File` and stored there
* improve the debugging function `Module.dumpInst`
* get rid of `Value.the_one_possible_value` in favor of a few more
specific values for different types. This is less buggy, one less
footgun.
* `Type.onePossibleValue` now returns a `?Value` instead of `bool`.
* codegen handles undefined values. `undef` is a new `MCValue` tag.
It uses 0xaa values depending on optimization mode. However
optimization mode does not yet support scope overrides.
* link.zig: move the `Options` field from `File.Elf` and `File.C` to
the base struct.
- fix the Tag enum to adhere to style conventions
* ZIR now supports emitting undefined values.
* Fix the logic of comptime math to properly compare against zero using
the `compareWithZero` function.
* implement sema for runtime deref, store pointer, coerce_to_ptr_elem,
and store
* identifiers support being lvalues, except for decls is still TODO
* codegen supports load, store, ref, alloc
* introduce more MCValue union tags to support pointers
* add load, ref, store typed IR instructions
* add Type.isVolatilePtr
Comment out non-x86_64 architectures for now in codegen.zig, because
they all have compile errors for their codepaths anyway, and it was
bloating the compilation speed and memory usage when stage1 tried to
build self-hosted. Here's the panic message:
"Backend architectures that don't have good support yet are commented
out, to improve compilation performance. If you are interested in one
of these other backends feel free to uncomment them. Eventually these
will be completed, but stage1 is slow and a memory hog."
This is a workaround to lower the time it takes to build self-hosted
with stage1 as well as use less memory. It should fix the CI.
Additionally:
* Add `single_mut_pointer` support to `Type`
* Trivial implementation of stack allocation in codegen.zig. It does
not deal with freeing yet, and it's missing the stack pointer
adjustment prologue.
* Add the `alloc` IR instruction and semantic analysis for `alloc` ZIR
instruction.
This makes sense from an organizational point of view, as explained by
this new doc comment at the top of the new file:
//! Semantic analysis of ZIR instructions.
//! This file operates on a `Module` instance, transforming untyped ZIR
//! instructions into semantically-analyzed IR instructions. It does type
//! checking, comptime control flow, and safety-check generation. This is the
//! the heart of the Zig compiler.
//! When deciding if something goes into this file or into Module, here is a
//! guiding principle: if it has to do with (untyped) ZIR instructions, it goes
//! here. If the analysis operates on typed IR instructions, it goes in Module.
Before:
4009 src-self-hosted/Module.zig
After:
2776 src-self-hosted/Module.zig
1128 src-self-hosted/zir_sema.zig
This should be sufficient to avoid the situation we have in stage1 where
ir.cpp is 32,516 lines.
