There are now very few stage1 cases remaining:
* `cases/compile_errors/stage1/obj/*` currently don't work correctly on
stage2. There are 6 of these, and most of them are probably fairly
simple to fix.
* `cases/compile_errors/async/*` and all remaining `safety/*` depend on
async; see #6025.
Resolves: #14849
Instead of using `zig test` to build a special version of the compiler
that runs all the test-cases, the zig build system is now used as much
as possible - all with the basic steps found in the standard library.
For incremental compilation tests (the ones that look like foo.0.zig,
foo.1.zig, foo.2.zig, etc.), a special version of the compiler is
compiled into a utility executable called "check-case" which checks
exactly one sequence of incremental updates in an independent
subprocess. Previously, all incremental and non-incremental test cases
were done in the same test runner process.
The compile error checking code is now simpler, but also a bit
rudimentary, and so it additionally makes sure that the actual compile
errors do not include *extra* messages, and it makes sure that the
actual compile errors output in the same order as expected. It is also
based on the "ends-with" property of each line rather than the previous
logic, which frankly I didn't want to touch with a ten-meter pole. The
compile error test cases have been updated to pass in light of these
differences.
Previously, 'error' mode with 0 compile errors was used to shoehorn in a
different kind of test-case - one that only checks if a piece of code
compiles without errors. Now there is a 'compile' mode of test-cases,
and 'error' must be only used when there are greater than 0 errors.
link test cases are updated to omit the target object format argument
when calling checkObject since that is no longer needed.
The test/stage2 directory is removed; the 2 files within are moved to be
directly in the test/ directory.
* Scan from line start when finding tag in tokenizer
This resolves a crash that can occur for invalid bytes like carriage
returns that are valid characters when not parsed from within literals.
There are potentially other edge cases this could resolve as well, as
the calling code for this function didn't account for any potential
'pending_invalid_tokens' that could be queued up by the tokenizer from
within another state.
* Fix carriage return crash in multiline string
Follow the guidance of #38:
> However CR directly before NL is interpreted as only a newline and not part of the multiline string. zig fmt will delete the CR.
Zig fmt already had code for deleting carriage returns, but would still
crash - now it no longer does so. Carriage returns encountered before
line-feeds are now appropriately removed on program compilation as well.
* Only accept carriage returns before line feeds
Previous commit was much less strict about this, this more closely
matches the desired spec of only allow CR characters in a CRLF pair, but
not otherwise.
* Fix CR being rejected when used as whitespace
Missed this comment from ziglang/zig-spec#83:
> CR used as whitespace, whether directly preceding NL or stray, is still unambiguously whitespace. It is accepted by the grammar and replaced by the canonical whitespace by zig fmt.
* Add tests for carriage return handling
The container we want to get the fields from might not be declared in the
same file as the block we are analyzing, so we should get the AST from
the decl's file instead.
* `-Dskip-compile-errors` is removed; `-Dskip-stage1` is added.
* Use `std.testing.allocator` instead of a new instance of GPA.
- Fix the memory leaks this revealed.
* Show the file name when it is not parsed correctly such as when the
manifest is missing.
- Better error messages when test files are not parsed correctly.
* Ignore unknown files such as swap files.
* Move logic from declarative file to the test harness implementation.
* Move stage1 tests to stage2 tests where appropriate.
Some cases had to stay behind, either because they required complex case
configuration that we don't support in independent files yet, or because
they have associated comments which we don't want to lose track of.
To make sure I didn't drop any tests in the process, I logged all
obj/test/exe test cases from a run of "zig build test" and compared
before/after this change.
All of the test cases match, with two exceptions:
- "use of comptime-known undefined function value" was deleted, since
it was a duplicate
- "slice sentinel mismatch" was renamed to "vector index out of
bounds", since it was incorrectly named
This brings two quality-of-life improvements for folks working on
compile error test cases:
- test cases can be added/changed without re-building Zig
- wrapping the source in a multi-line string literal is not necessary
I decided to keep things as simple as possible for this initial
implementation. The test "manifest" is a contiguous comment block at the
end of the test file:
1. The first line is the test case name
2. The second line is a blank comment
2. The following lines are expected errors
Here's an example:
```zig
const U = union(enum(u2)) {
A: u8,
B: u8,
C: u8,
D: u8,
E: u8,
};
export fn entry() void {
_ = U{ .E = 1 };
}
// union with too small explicit unsigned tag type
//
// tmp.zig:1:22: error: specified integer tag type cannot represent every field
// tmp.zig:1:22: note: type u2 cannot fit values in range 0...4
```
The mode of the test (obj/exe/test), as well as the target
(stage1/stage2) is determined based on the directory containing the
test.
We'll probably eventually want to support embedding this information
in the test files themselves, similar to the arocc test runner, but
that enhancement can be tackled later.
The runtime behavior allowed this in both stage1 and stage2, but stage1
fails with index out of bounds during comptime. This behavior makes
sense to support, and comptime behavior should match runtime behavior. I
implement this fix only in stage2.
If a '(' is found where the continue expression was expected and it is
on the same line as the previous token issue an error about missing
colon before the continue expression.
For some errors if the found token is not on the same line as
the previous token, point to the end of the previous token.
This usually results in more helpful errors.
The main problem was that the loop body was treated as an expression
that was one of the peer result values of a loop, when in reality the
loop body is noreturn and only the `break` operands are the result
values of loops.
This was solved by introducing an override that prevents rvalue() from
emitting a store to result location instruction for loop bodies.
An orthogonal change also included in this commit is switching
`elem_val` index expressions to using `coerced_ty` and doing the
coercion to `usize` inside `Sema`, resulting in smaller ZIR (since the
cast becomes implied).
I also changed the break operand expression to use `reachableExpr`,
introducing a new compile error for double break.
This makes a few more behavior tests pass for `while` and `for` loops.
The previous commit (38b2d62092) regressed
the compile error test case for when doing saturating shift left of a
comptime-known negative RHS.
This commit additionally fixes the error for regular shifts in addition
to saturating shifts.
* stage1: change the `@typeName` of `@TypeOf(undefined)`,
`@TypeOf(null)`, and `@TypeOf(.foo)` to match stage2.
* move passing behavior tests to the passing-for-stage2 section.
After extern enums were removed, stage1 was left in an incorrect state
of checking for `extern enum` for exported enums. This commit fixes it
to look for an explicit integer tag type instead, and adds test coverage
for the compile error case as well as the success case.
closes#9498
When adding test coverage, I noticed an inconsistency in which source
location the compile error was pointing to for `@embedFile` errors vs
`@import` errors. They now both point to the same place, the string
operand.
closes#9404closes#9939
