Instead of Module setting up the root_scope with the root source file,
instead, Module relies on the package table graph being set up properly,
and inside `update()`, it does the equivalent of `_ = @import("std");`.
This, in term, imports start.zig, which has the logic to call main (or
not). `Module` no longer has `root_scope` - the root source file is no
longer special, it's just in the package table mapped to "root".
I also went ahead and implemented proper detection of updated files.
mtime, inode, size, and source hash are kept in `Scope.File`.
During an update, iterate over `import_table` and stat each file to find
out which ones are updated.
The source hash is redundant with the source hash used by the struct
decl that corresponds to the file, so it should be removed in a future
commit before merging the branch.
* AstGen: add "previously declared here" notes for variables shadowing
decls.
* Parse imports as structs. Module now calls `AstGen.structDeclInner`,
which is called by `AstGen.containerDecl`.
- `importFile` is a bit kludgy with how it handles the top level Decl
that kinda gets merged into the struct decl at the end of the
function. Be on the look out for bugs related to that as well as
possibly cleaner ways to implement this.
* Module: factor out lookupDeclName into lookupIdentifier and lookupNa
* Rename `Scope.Container` to `Scope.Namespace`.
* Delete some dead code.
This branch won't work until `usingnamespace` is implemented because it
relies on `@import("builtin").OutputMode` and `OutputMode` comes from a
`usingnamespace`.
This branch adds "builtin" and "std" to the import table when using the
self-hosted backend.
"builtin" gains one additional item:
```
pub const zig_is_stage2 = true; // false when using stage1 backend
```
This allows the std lib to do conditional compilation based on detecting
which backend is being used. This will be removed from builtin as soon
as self-hosted catches up to feature parity with stage1.
Keep a sharp eye out - people are going to be tempted to abuse this.
The general rule of thumb is do not use `builtin.zig_is_stage2`. However
this commit breaks the rule so that we can gain limited start.zig support
as we incrementally improve the self-hosted compiler.
This commit also implements `fullyQualifiedNameHash` and related
functionality, which effectively puts all Decls in their proper
namespaces. `fullyQualifiedName` is not yet implemented.
Stop printing "todo" log messages for test decls unless we are in test
mode.
Add "previous definition here" error notes for Decl name collisions.
This commit does not bring us yet to a newly passing test case.
Here's what I'm working towards:
```zig
const std = @import("std");
export fn main() c_int {
const a = std.fs.base64_alphabet[0];
return a - 'A';
}
```
Current output:
```
$ ./zig-cache/bin/zig build-exe test.zig
test.zig:3:1: error: TODO implement more analyze elemptr
zig-cache/lib/zig/std/start.zig:38:46: error: TODO implement structInitExpr ty
```
So the next steps are clear:
* Sema: improve elemptr
* AstGen: implement structInitExpr
This refactor inserts an offset table into wasm's data section
where each offset points to the actual data region.
This means we can keep offset indexes consistant and do not
have to perform any computer to determine where in the data section
something like a static string exists. Instead during runtime
it will load the data offset onto the stack.
The following code caused an assertion to be hit:
```
pub fn main() void {
var e: anyerror!c_int = error.Foo;
const i = e catch 69;
assert(69 - i == 0);
}
```
There were several problems, all fixed:
* AstGen was storing field names as references to the original
source code bytes. However, that data would be destroyed when the
source file is updated. Now, it correctly stores the field names in
the Decl arena for the enum. The same fix applies to error set field
names.
* Sema was missing a memset inside `analyzeSwitch`, leaving the "seen
enum fields" array with undefined memory. Now that they are all
properly set to null, the validation works.
* Moved the "enum declared here" note to the end. It looked weird
interrupting the notes for which enum values were missing.
Before, incremental compilation would crash when trying to emit compile
errors for the update after introducing a parse error.
Parse errors are handled by not invalidating any existing semantic
analysis. However, only the parse error must be reported, with all the
other errors suppressed. Once the parse error is fixed, the new file can
be treated as an update to the previously-succeeded update.
* `analyzeContainer` now has an `outdated_decls` set as well as
`deleted_decls`. Instead of queuing up outdated Decls for re-analysis
right away, they are added to this new set. When processing the
`deleted_decls` set, we remove deleted Decls from the
`outdated_decls` set, to avoid deleted Decl pointers from being in
the work_queue. Only after processing the deleted decls do we add
analyze_decl work items to the queue.
* Module.deletion_set is now an `AutoArrayHashMap` rather than `ArrayList`.
`declareDeclDependency` will now remove a Decl from it as appropriate.
When processing the `deletion_set` in `Compilation.performAllTheWork`,
it now assumes all Decl in the set are to be deleted.
* Fix crash when handling parse errors. Currently we unload the
`ast.Tree` if any parse errors occur. Previously the code emitted a
LazySrcLoc pointing to a token index, but then when we try to resolve
the token index to a byte offset to create a compile error message,
the ast.Tree` would be unloaded. Now we use
`LazySrcLoc.byte_abs` instead of `token_abs` so the error message can
be created even with the `ast.Tree` unloaded.
Together, these changes solve a crash that happened with incremental
compilation when Decls were added and removed in some combinations.
A simple enum is an enum which has an automatic integer tag type,
all tag values automatically assigned, and no top level declarations.
Such enums are created directly in AstGen and shared by all the
generic/comptime instantiations of the surrounding ZIR code. This
commit implements, but does not yet add any test cases for, simple enums.
A full enum is an enum for which any of the above conditions are not
true. Full enums are created in Sema, and therefore will create a unique
type per generic/comptime instantiation. This commit does not implement
full enums. However the `enum_decl_nonexhaustive` ZIR instruction is
added and the respective Type functions are filled out.
This commit makes an improvement to ZIR code, removing the decls array
and removing the decl_map from AstGen. Instead, decl_ref and
decl_val ZIR instructions index into the `owner_decl.dependencies`
ArrayHashMap. We already need this dependencies array for incremental
compilation purposes, and so repurposing it to also use it for ZIR decl
indexes makes for efficient memory usage.
Similarly, this commit fixes up incorrect memory management by removing
the `const` ZIR instruction. The two places it was used stored memory in
the AstGen arena, which may get freed after Sema. Now it properly sets
up a new anonymous Decl for error sets and uses a normal decl_val
instruction.
The other usage of `const` ZIR instruction was float literals. These are
now changed to use `float` ZIR instruction when the value fits inside
`zir.Inst.Data` and `float128` otherwise.
AstGen + Sema: implement int_to_enum and enum_to_int. No tests yet; I expect to
have to make some fixes before they will pass tests. Will do that in the
branch before merging.
AstGen: fix struct astgen incorrectly counting decls as fields.
Type/Value: give up on trying to exhaustively list every tag all the
time. This makes the file more manageable. Also found a bug with
i128/u128 this way, since the name of the function was more obvious when
looking at the tag values.
Type: implement abiAlignment and abiSize for structs. This will need to
get more sophisticated at some point, but for now it is progress.
Value: add new `enum_field_index` tag.
Value: add hash_u32, needed when using ArrayHashMap.
Introduce `ResultLoc.none_or_ref` which is used by field access
expressions to avoid unnecessary loads when the field access itself
will do the load. This turns:
```zig
p.y - p.x - p.x
```
from
```zir
%14 = load(%4) node_offset:8:12
%15 = field_val(%14, "y") node_offset:8:13
%16 = load(%4) node_offset:8:18
%17 = field_val(%16, "x") node_offset:8:19
%18 = sub(%15, %17) node_offset:8:16
%19 = load(%4) node_offset:8:24
%20 = field_val(%19, "x") node_offset:8:25
```
to
```zir
%14 = field_val(%4, "y") node_offset:8:13
%15 = field_val(%4, "x") node_offset:8:19
%16 = sub(%14, %15) node_offset:8:16
%17 = field_val(%4, "x") node_offset:8:25
```
Much more compact. This requires `Sema.zirFieldVal` to support both
pointers and non-pointers.
C backend: Implement typedefs for struct types, as well as the following
TZIR instructions:
* mul
* mulwrap
* addwrap
* subwrap
* ref
* struct_field_ptr
Note that add, addwrap, sub, subwrap, mul, mulwrap instructions are all
incorrect currently and need to be updated to properly handle wrapping
and non wrapping for signed and unsigned.
C backend: change indentation delta to 1, to make the output smaller and
to process fewer bytes.
I promise I will add a test case as soon as I fix those warnings that
are being printed for my test case.
New ZIR instructions:
* struct_decl_packed
* struct_decl_extern
New TZIR instruction: struct_field_ptr
Introduce `Module.Struct`. It uses `Value` to store default values and
abi alignments.
Implemented Sema.analyzeStructFieldPtr and zirStructDecl.
Some stuff I changed from `@panic("TODO")` to `log.warn("TODO")`.
It's becoming more clear that we need the lazy value mechanism soon;
Type is becoming unruly, and some of these functions have too much logic
given that they don't have any context for memory management or error
reporting.
GenZir struct now has rl_ty_inst field which tracks the result location
type (if any) a block expects all of its results to be coerced to.
Remove a redundant coercion on const local initialization with a
specified type.
Switch expressions, during elision of store_to_block_ptr instructions,
now re-purpose them to be type coercion when the block has a type in the
result location.
The logic for putting ranges into the else prong is moved from AstGen to
Sema. However, logic to emit multi-items the same as single-items cannot
be done until TZIR supports mapping multiple items to the same block of
code. This will be simple to represent when we do the upcoming TZIR memory
layout changes.
Not yet implemented in this commit is the validation of duplicate
values. The trick is going to be emitting error messages with accurate
source locations, without adding extra source nodes to the ZIR
switch instruction.
This will be done by computing the respective AST node based on the
switch node (which we do have available), only when a compile error
occurs and we need to know the source location to attach the message to.
Here's what I think the ZIR should be. AstGen is not yet implemented to
match this, and the main implementation of analyzeSwitch in Sema is not
yet implemented to match it either.
Here are some example byte size reductions from master branch, with the
ZIR memory layout from this commit:
```
switch (foo) {
a => 1,
b => 2,
c => 3,
d => 4,
}
```
184 bytes (master) => 40 bytes (this branch)
```
switch (foo) {
a, b => 1,
c..d, e, f => 2,
g => 3,
else => 4,
}
```
240 bytes (master) => 80 bytes (this branch)
* zir.Code: introduce a decls array. This is so that `decl_val` and
`decl_ref` instructions can refer to a Decl with a u32 and therefore
they can also store a source location. This is needed for proper
compile error reporting.
* astgen uses a hash map to avoid redundantly adding a Decl to the
decls array.
* fixed reporting "instruction illegal outside function body" instead
of the desired message "unable to resolve comptime value".
* astgen skips emitting dbg_stmt instructions in comptime scopes.
* astgen has some logic to avoid adding unnecessary type coercion
instructions for common values.
Introduce "inline" variants of ZIR tags:
* block => block_inline
* repeat => repeat_inline
* break => break_inline
* condbr => condbr_inline
The inline variants perform control flow at compile-time, and they
utilize the return value of `Sema.analyzeBody`.
`analyzeBody` now returns an Index, not a Ref, which is the ZIR index of
a break instruction. This effectively communicates both the intended
break target block as well as the operand, allowing parent blocks to
find out whether they, in turn, should return the break instruction up the
call stack, or accept the operand as the block's result and continue
analyzing instructions in the block.
Additionally:
* removed the deprecated ZIR tag `block_comptime`.
* removed `break_void_node` so that all break instructions use the same Data.
* zir.Code: remove the `root_start` and `root_len` fields. There is now
implied to be a block at index 0 for the root body. This is so that
`break_inline` has something to point at and we no longer need the
special instruction `break_flat`.
* implement source location byteOffset() for .node_offset_if_cond
.node_offset_for_cond is probably redundant and can be deleted.
We don't have `comptime var` supported yet, so this commit adds a test
that at least makes sure the condition is required to be comptime known
for `inline while`.
* Module.addBreak and addBreakVoid return zir.Inst.Index not Ref
because Index is the simpler type and we never need a Ref for these.
* astgen: make noreturn stuff return the unreachable_value and avoid
unnecessary calls to rvalue()
* breakExpr: avoid unnecessary access into the tokens array
* breakExpr: fix incorrect `@intCast` (previously this unsafely
casted an Index to a Ref)
* Introduce helper functions on Module.WipZirCode and zir.Code
* Move some logic around
* re-introduce ref_start_index
* prefer usize for local variables + `@intCast` at the end.
Empirically this is easier to optimize.
* Avoid using mem.{bytesAsSlice,sliceAsBytes} because it incurs an
unnecessary multiplication/division which may cause problems for the
optimizer.
* Use a regular enum, not packed, for `Ref`. Memory layout is
guaranteed for enums which specify their tag type. Packed enums have
ABI alignment of 1 byte which is too small.
This provides us greatly increased type safety and prevents the common
mistake of using a zir.Inst.Ref where a zir.Inst.Index was expected or
vice-versa. It also increases the ergonomics of using the typed values
which can be directly referenced with a Ref over the previous zir.Const
approach.
The main pain point is casting between a []Ref and []u32, which could be
alleviated in the future with a new std.mem function.
* comment out the failing stage2 test cases
(so that we can uncomment the ones that are newly passing with
further commits)
* Sema: implement negate, negatewrap
* astgen: implement field access, multiline string literals, and
character literals
* Module: when resolving an AST node into a byte offset, use the
main_tokens array, not the firstToken function
