In the docker container, the ipv6 loopback address is not set,
and connect("::1") has to return ENEADDRNOTAVAIL in this case.
Without this fix, it returns EHOSTUNREACH.
PiperOrigin-RevId: 340002915
#4641 fixed the PHP runtime test ext/standard/tests/network/bug20134.phpt.
We should start testing it again.
Also excluded another flaky test. Seems like a test bug.
PiperOrigin-RevId: 339475716
The python:test_subprocess enumerates all possible file descriptors and fails
by timeout if the limit is too high.
There is a know thing about docker that it sets this limit to 1M by default,
but on native linux, this limit will be between 1K to 32K.
PiperOrigin-RevId: 338197239
bhaskerh@ fixed a bunch of the EADDRINUSE flakes in #3662 so we should
unexclude them.
I have also tested other flaky tests on this list and removed those that do
not flake anymore.
PiperOrigin-RevId: 338158545
Regarding ThreadCpuTimeArray.java: The test starts 10 threads, each of which
does some computation, then blocks. When all threads are blocked, the test
sleeps for 200ms, then checks that less than 100ns of CPU time in userspace
elapse over the course of the sleep; AFAICT, the 100ns of slop is because a
thread indicates that it's in the WAITING state before it actually blocks, and
because signals can cause threads to be temporarily woken. gVisor's CPU clocks
have a granularity of 10ms (the interval of Kernel.cpuClockTicker is
//pkg/abi/linux.ClockTick), so a single tick pushes the test over the
threshold.
PiperOrigin-RevId: 333830287
Added a README describing what these tests are, how they work and how to run
them locally. Also reorganized the exclude files into a directory.
PiperOrigin-RevId: 332079697
These tests print disk_free_space()/disk_total_space() and expect the printed
result to be an integer (despite the fact that both the documented and returned
type is float). After cl/297213789, free/total disk space on tmpfs is
sufficiently large that PHP prints the result in scientific notation instead:
========DIFF========
012+ float(9.2233720368548E+18)
013+ float(9.2233720368548E+18)
012- float(%d)
013- float(%d)
========DONE========
FAIL disk_total_space() and disk_free_space() tests [ext/standard/tests/file/disk.phpt]
PiperOrigin-RevId: 328349906
test_httplib was un-excluded in cl/325137657. Exclude it again.
test_httplib fails consistently when running in a kokoro job. Could not repro
with Docker.
There are some java tests that pass locally on Docker but fail when running
within a kokoro job.
Also make shard count a multiple of 8 to minimize idle cores and minimize test
time.
PiperOrigin-RevId: 325267071
- un-exclude passing tests to increase testing surface.
- create/update bugs for tests that fail on runsc but pass on runc.
PiperOrigin-RevId: 325137657
- Exclude flaky tests.
- Bump timeout.
- Un-exclude passing tests to increase testing surface.
- Create/Update bugs for tests that pass on runc but fail on runsc.
PiperOrigin-RevId: 324830840
- Added a bunch of helpful options which help in speeding up the test and
providing useful output.
- Unexcluded passing tests and updated bugs. Excluded tests which were failing.
- Increased the batch size for java tests so that we can take advantage of
the shared JVMs.
The running time of the tests decreased from 3+ hours (I don't know the exact
running time because this test has always timed out after 3 hours) to 1 hour
15 minutes. We can reliably run this a CI kokoro job.
PiperOrigin-RevId: 324301503
- Unexported some passing tests. This will increase the testing surface and
will be especially helpful when this is enabled for vfs2.
- Run tool tests with -v (verbose output). We only print the output when a test
fails so this should not clutter the output.
- Run tool tests with "-no-rebuild" flag.
- Surround test name with appropriate regex, i.e. ^testname$. This will ensure
that only that test is run. Earlier running go_test:os would also run
go_test:os/exec, go_test:os/signal, go_test:os/user. This should help speed
up the tests as we do not run the same test multiple times anymore.
- Updated bugs.
Updates #3191
PiperOrigin-RevId: 324028878
Earlier we were docker exec-ing each test at a time. However invoking the test
framework has a fixed overhead which made it infeasible to make the runtime
tests run as presubmits. This change now executes tests in batches of 50 (can
be altered). This really speeds up testing process.
With this change, the following tests can be run in reasonable times:
- Go
- Nodejs
- Php
- Python
PiperOrigin-RevId: 320763916
Fixed an issue with the runtime test runner which enables us to run tests in
shards. We had to touch the status file as indicated by an env var.
PiperOrigin-RevId: 320236205
This change moves all Docker images to a standard location, and abstracts the
build process so that they can be maintained in an automated fashion. This also
allows the images to be architecture-independent.
All images will now be referred to by the test framework via the canonical
`gvisor.dev/images/<name>`, where `<name>` is a function of the path within the
source tree.
In a subsequent change, continuous integration will be added so that the images
will always be correct and available locally.
In the end, using `bazel` for Docker containers is simply not possible. Given
that we already have the need to use `make` with the base container (for
Docker), we extend this approach to get more flexibility.
This change also adds a self-documenting and powerful Makefile that is intended
to replace the collection of scripts in scripts. Canonical (self-documenting)
targets can be added here for targets that understand which images need to be
loaded and/or built.
PiperOrigin-RevId: 308322438
This change adds a layer of abstraction around the internal Docker APIs,
and eliminates all direct dependencies on Dockerfiles in the infrastructure.
A subsequent change will automated the generation of local images (with
efficient caching). Note that this change drops the use of bazel container
rules, as that experiment does not seem to be viable.
PiperOrigin-RevId: 308095430
* Tests are picked for a shard differently. It now picks one test from each
block, instead of picking the whole block. This makes the same kind of tests
spreads across different shards.
* Reduce the number of connect() calls in TCPListenClose.
PiperOrigin-RevId: 293019281
This was intended behavior per the README, but running tests without the --test
flag caused an error. Users can now omit the --test flag to run every test for a
runtime.
PiperOrigin-RevId: 280522025
Ayush Ranjan
Andrei Vagin
Jamie Liu
Adin Scannell
Dean Deng
Zach Koopmans
Fabricio Voznika
Michael Pratt
gVisor bot
Ting-Yu Wang
Kevin Krakauer
Nicolas Lacasse