By requiring the source file to be null-terminated, we avoid extra
branching while simplifying the logic at the same time.
Running ast-check on a large zig source file (udivmodti4_test.zig),
master branch compared to this commit:
* 4% faster wall clock
* 7% fewer cache misses
* 1% fewer branches
The motivation for this commit is that there exists source files which
produce ast-check errors, but crash stage1 or otherwise trigger stage1
bugs. Previously to this commit, Zig would run AstGen, collect the
compile errors, run stage1, report stage1 compile errors and exit if
any, and then report AstGen compile errors.
The main change in this commit is to report AstGen errors prior to
invoking stage1, and in fact if any AstGen errors occur, do not invoke
stage1 at all.
This caused most of the compile error tests to fail due to things such
as unused local variables and mismatched stage1/stage2 error messages.
It was taking a long time to update the test cases one-by-one, so I
took this opportunity to unify the stage1 and stage2 testing harness,
specifically with regards to compile errors. In this way we can start
keeping track of which tests pass for 1, 2, or both.
`zig build test-compile-errors` no longer works; it is now integrated
into `zig build test-stage2`.
This is one step closer to executing compile error tests in parallel; in
fact the ThreadPool object is already in scope.
There are some cases where the stage1 compile errors were actually
better; those are left failing in this commit, to be addressed in a
follow-up commit.
Other changes in this commit:
* build.zig: improve support for -Dstage1 used with the test step.
* AstGen: minor cosmetic changes to error messages.
* stage2: add -fstage1 and -fno-stage1 flags. This now allows one to
download a binary of the zig compiler and use the llvm backend of
self-hosted. This was also needed for hooking up the test harness.
However, I realized that stage1 calls exit() and also has memory
leaks, so had to complicate the test harness by not using this flag
after all and instead invoking as a child process.
- These CLI flags will disappear once we start shipping the
self-hosted compiler as the main compiler. Until then, they can be
used to try out the work-in-progress stage2.
* stage2: select the LLVM backend by default for release modes, as long
as the target architecture is supported by LLVM.
* test harness: support setting the optimize mode
* AstGen: implement "unreachable code" error for blocks. This works at
the statement level.
* stage1: remove the "unreachable code" error implementation, which
means removing the `is_gen` field from IrInstSrc. This is one small
step towards a smaller memory footprint for stage1. The benefits
won't be realized until a future commit because this flag took
advantage of padding.
There may be a regression here with "union has no associated enum"
error, and there is a regression with the following code:
```zig
const a = noreturn;
```
A future commit will address these regressions.
When a floating-point value with no fractional part is shoved into an
integer type we must check whether it fits or not before calling
`@floatToInt` as the builtin panics in case of overflow.
Catch the error and bubble it up to the caller.
* Add command line help for "-mexec-model"
* Define WasmExecModel enum in std.builtin.
* Drop the support for the old crt1.o in favor of crt1-command.o
Signed-off-by: Takeshi Yoneda <takeshi@tetrate.io>
Update to accomodate the differences in Windows, which is now advisory
file locking, and include details about which operating systems have
atomic locking flags.
When working with durations it often makes sense to use signed integers
and allow negative durations, and there is currently no nice way to
format these in std.fmt. This patch adds a simple wrapper for the
existing fmtDurtion to fit this need.
With this change zig ld can link with dynamic libraries
contained within a fat/universal file that had multiple
seperate binaries embedded within it for multi-arch
support (in macOS).
Whilst zig can still only create single-architecture
executables - the ability to link with fat libraries is
useful for cases where they are the easiest (or only)
option to link against.
Translate enum types as the underlying integer type. Translate enum constants
as top-level integer constants of the correct type (which does not necessarily
match the enum integer type).
If an enum constant's type cannot be translated for some reason, omit it.
See discussion https://github.com/ziglang/zig/issues/2115#issuecomment-827968279Fixes#9153
Previously the fd parameter was ignored and so the result would not get
populated. Now it passes the fd pointer to the inline assembly so that
the results can be observed.
* Remove parser error on double ampersand
* Add failing test for double ampersand case
* Add error when encountering double ampersand in AstGen
"Bit and" operator should not make sense when one of its operands
is an address.
* Check that 2 ampersands are adjacent to each other in source string
* Remove cases of unused variables in tests
* Avoid emitting the copy_file_range symbol at all to prevent link-time
errors.
* Fix a bug in the check logic, the has_copy_file_range_syscall was
set to the wrong value in case of ENOSYS
* If link_libc is true don't fall-back to the raw syscall approach,
there's no policy about what to do in this case but let's follow what
the other impls do.
Fixes#9146
Previous to #7082, users could overwrite PATH_MAX in the root file to support std.os.toPosixPath, permitting the "bring your own operating system" layer to implement the POSIX API for opening files. Unfortunately that is no longer the case.
This commit intends to fix what is arguably a regression from 0.7 in a way that doesn't break any code targeting 0.8.0, making it suitable to be included in a 0.8 patch release.
However in a future release that permits breaking changes, I am of the opinion that it would be beneficial to overwrite the value, even for "supported" operating systems. Same for all the other POSIX/BYOOS functions and values. However this is beyond the scope of this commit. Further discussion of this will be made into an issue in due time.
* Don't skip the TLS initialization (Fixes#9083)
* Add a test case where a PIE program is built and run
* Refactor the common initialization code in the Linux startup
sequence.
This finishes LemonBoy's Draft PR ziglang#6750. It updates ChildProcess to collect the output from stdout/stderr asynchronously using Overlapped IO and named pipes.
Keep polling until there are enough open handles, if the child process
terminates closing the handles or explicitly closes them we just quit
polling and wait for the process handle to signal the termination
condition.
Reading stdin&stderr at different times may lead to nasty deadlocks (eg.
when stdout is read before stderr and the child process doesn't write
anything onto stdout).
Implement a polling mechanism to make sure this won't happen: we read
data from stderr/stdout as it becomes ready and then it's copied into an
ArrayList provided by the user, avoiding any kind of blocking read.
Lakemont has no x86, no MMX, no SSE and no way of handling any fp-math. In theory LLVM is able to implicitly use the soft-float emulation library calls to legalize any such operation but, given Zig's use of many non-standard features, sometimes we hit a weak spot in the X86 codegen backend.
Consider this as a work-around for this LLVM problem, fixing the problem in LLVM is not so high in my todo list as the target is pretty niche and Intel axed it in '19.
(Commit message by @LemonBoy)
