Ensures that if an assignment statement is the sole statement within a
C if statement, for loop, do loop, or do while loop, then when translated
it resides within a block, even though it does not in the original C.
Fixes the following invalid translation:
`if (1) if (1) 2;` -> `if (true) if (true) _ = @as(c_int, 2);`
To this:
```zig
if (true) if (true) {
_ = @as(c_int, 2);
};
```
Fixes#8159
Now that we ship our own linker for MachO by default in both stage1
and stage2, we need a way to enable logs for verbose debugging.
This commit adds `ZIG_ENABLE_LOGGING` cmake option which is equivalent
to stage2's `-Dlog` flag.
To enable it when building stage1 with cmake, add:
```
cmake .. -DZIG_ENABLE_LOGGING=on
```
move a boolean field to be represented implicitly with the enum tag.
Just borrowing one of the many strategies of stage2.
This simple change took the peak mem usage from std lib tests on
my machine from 8.21 GiB to 8.11 GiB.
* translate-c: Use [N:0] arrays when initializer is a string literal
Translate incomplete arrays as [N:0] when initialized by a string literal.
This preserves a bit more of the type information from the original C program.
Fixes#8215
Now reading a source file and decoding it from UTF-16LE to UTF-8 is
done in a single function. Error messages are improved, and an error is
emitted when the source file has a BOM not supported (UTF-16BE, UTF-32).
Please note that the BOM of UTF-32 is composed of the same bytes as the
BOM of UTF-16 followed by a null character. Therefore a source file in
UTF-16LE starting with a null byte will be interpreted as an UTF-32, and
rejeted because of an invalid format. In pratice this is not a problem,
as the code would have been rejected later anyway because of the null
character.
Fixes#2820
After reading the source code, the first two bytes are inspected, and
if they correspond to a UTF-16 BOM in little-endian order, the source
code is converted to UTF-8.
Luuk de Gram