-
Notifications
You must be signed in to change notification settings - Fork 18
Note testsuite
The rust test suite has several sets of tests for different purposes. As the compiler is built several times, over multiple stages, the tests can genarally be run against each stage of the build. Under linux, all tests are run under valgrind if it is installed, unless the build was configured with --disable-valgrind or CFG_DISABLE_VALGRIND=1 is provided on the command line.
- Run the test suite:
make check
. This runs all stage2 tests. This is the criteria for a successful build. - Run only xfailed (ignored) tests:
make check CHECK_XFAILS=1
- Run with verbose output:
make check VERBOSE=1
. This will print useful information like the exact commands being run. - Run without valgrind:
make check CFG_DISABLE_VALGRIND=1
- Run a specific test:
make check TESTNAME=[...]
- Note that while this will run only tests matching the given pattern, it will still execute all test runners - most of them will just not execute any tests. For more precise control, call make on one of the targets below. - Run without parallelism:
RUST_THREADS=1 make check
- This can make it easier to understand failure output.
These options can be combined. For instance, make check CHECK_XFAILS=1 TESTNAME=test/run-pass/foobar.rs
runs the xfailed test foobar.rs
in the run-pass
directory.
These are tests of the compiler as a whole. Each test is a source file or source crate over which the compiler is run in some configuration and the resulting executable run. They typically have a main
function that takes no arguments and may have directives that instruct the test runner how to run the test.
The test runner for these tests is at src/test/compiletest and is compiled to test/compiletest.stage[N].
A typical test might look like:
// Regression test for issue #XXX
fn main() {
let a: bool = 10; //! ERROR mismatched types
}
There are four different modes for compile tests. Each test is run under one or more modes:
- compile-fail - The test should fail to compile. Must include at least one expected error.
- run-fail - The test should compile but fail to run. Must include at least one error-pattern directive.
- run-pass - The test should compile and run successfully
- pretty - The test should round-trip through the pretty-printer and then compile successfully
Valid directives include:
-
compile-flags:[...]
- Additional arguments to pass to the compiler -
pp-exact
- The test should pretty-print exactly as written -
pp-exact:[filename]
- The pretty-printed test should match the example infilename
-
xfail-test
- Test is broken, don't run it -
xfail-fast
- Don't run as part of check-fast, a special win32 test runner (some tests don't work with it) -
xfail-pretty
- Test doesn't pretty-print correcty -
error-pattern:[...]
- A message that should be expected on standard out. If multiple patterns are provided then they must all be matched, in order (Note: error-patterns are somewhat deprecated, see the section on Expected Errors below).
There are five directories containing compile tests, living in the src/tests directory:
- run-pass - Tests that are expected to compile and run successfully. Also used for pretty-print testing.
- run-fail - Tests that are expected compile but fail when run. Also used for pretty-print testing.
- compile-fail - Tests that are expected not to compile
- bench - Benchmarks and miscellaneous snippets of code that are expected to compile and run successfully. Also used for pretty-print testing.
- pretty - Pretty-print tests
And finally, build targets:
- check-stage[N]-rpass - The tests in the run-pass directory, in run-pass mode
- check-stage[N]-rfail - The tests in the run-fail-directory, in run-fail mode
- check-stage[N]-cfail - The tests in the compile-fail directory, in compile-fail mode
- check-stage[N]-bench - The tests in the bench directory, in run-pass mode
- check-stage[N]-pretty - All the pretty-print tests
- check-stage[N]-pretty-rpass - The tests in the run-pass directory, in pretty mode
- check-stage[N]-pretty-rfail - The tests in the run-fail directory, in pretty mode
- check-stage[N]-pretty-bench - The tests in the bench directory, in pretty mode
- check-stage[N]-pretty-pretty - The tests in the pretty directory, in pretty mode
When writing a compile-fail test, you must specify at least one
expected error or warning message. The preferred way to do this is to
place a comment with the form //! ERROR msg
or //! WARNING msg
on
the line where the error or warning is expected to occur. You may
have as many of these comments as you like. The test harness will
verify that the compiler reports precisely the errors/warnings that are
specified, no more and no less. An example of using the error/warning
messages is:
// Regression test for issue #XXX
fn main() {
let a: bool = 10; //! ERROR mismatched types
log (debug, b);
}
In fact, this test would fail, because there are two errors: the type
mismatch and the undefined variable b
.
Sometimes it is not possible or not convenient to place the //!
comment on precisely the line where the error occurs. For those cases,
you may make a comment of the form //!^
where the carrot ^
indicates that the error is expected to appear on the line above. You
may have as many carrots as you like, so //!^^^ ERROR foo
indicates
that the error message foo
is expected to be reported 3 lines above
the comment. We could therefore correct the above test like so:
// Regression test for issue #XXX
fn main() {
let a: bool = 10; //! ERROR mismatched types
log (debug, b);
//!^ ERROR undefined variable `b`
}
The older technique for specify error messages is to use an
error-pattern
directive. These directives are placed at the top of
the file and each message found in an error-pattern
directive must
appear in the output. Using error comments is preferred, however,
because it is a more thorough test: (a) it verifies that the error is
reported on the expected line number and (b) it verifies that no
additional errors or warnings are reported.
- Running the run-pass tests for stage1:
make check-stage1-rpass
- Running a specific compile-fail test:
make check-stage2-cfail TESTNAME=type-mismatch
- Finding the command to compile a failing test:
make check-stage1-rpass TESTNAME=hello VERBOSE=1
- Running xfailed tests:
make check-stage1-rpass CHECK_XFAILS=1
These are unit tests against rustc internals. They are part of the rustc crate, and use the standard Rust test runner. Their test runner is compiled to test/rustctest.stage[N]
Unit test for rustc go near the code they test and should be wrapped in a conditionally-compiled test module like so
#[cfg(test)]
mod test {
#[test]
fn resolve_fn_types() {
...
}
}
Note that currently tests need to be visible at the top level of a crate in order to run.
- Running the tests:
make check-stage[N]-rustc
- Running a specific test:
make check-stage[N]-rustc TESTNAME=[test]
- Running ignored tests:
make check-stage[N]-rustc CHECK_XFAILS=1
Tests of the standard library live in a separate stdtest crate, under the test/stdtest directory. The compiled test runner lives at test/stdtest.stage[N]
- Running the tests:
make check-stage[N]-std
- Running a specific test:
make check-stage[N]-std TESTNAME=[test]
- Running ignored tests:
make check-stage[N]-std CHECK_XFAILS=1
Because Windows has slow process spawning running make check
on that platform can take a long time. For this reason we have a make check-fast
target that the Windows build servers run to keep the cycle time down. This is a special test runner that is built by combining all the run-pass tests into a single library. It is created by the src/etc/combine-tests.py script.