Re: [llvm-dev] Timeout tests timing out
Sorry I haven't had the time to dig into the issue but it looks someone else already fixed it :) On Thu, 3 Dec 2020 at 21:00, David Blaikie wrote: > > This'll hopefully be addressed by https://reviews.llvm.org/D92563 > > On Mon, Nov 30, 2020 at 6:28 AM Diana Picus wrote: >> >> Ping again. We're seeing this on several aarch64 bots, what can we do about >> it? >> >> On Mon, 23 Nov 2020 at 21:19, David Blaikie via llvm-dev >> wrote: >>> >>> Ping on this - Dan, any chance you could take a look here? >>> >>> On Mon, Nov 9, 2020 at 1:48 PM Arthur Eubanks wrote: >>> > >>> > Another case: http://lab.llvm.org:8011/#/builders/43/builds/810 >>> > shtest-timeout.py seems to be fairly flaky on the >>> > clang-cmake-aarch64-quick bot: http://lab.llvm.org:8011/#/builders/43, I >>> > get notifications from it fairly often >>> > >>> > On Thu, Oct 22, 2020 at 7:15 PM David Blaikie via llvm-dev >>> > wrote: >>> >> >>> >> Looks like there might still be some issues with the timeout tests? >>> >> http://lab.llvm.org:8011/#/builders/126/builds/226/steps/13/logs/FAIL__lit___shtest-timeout_py >>> >> >>> >> On Sun, Oct 4, 2020 at 2:44 PM Dan Liew wrote: >>> >>> >>> >>> > > One thing we could do to remove fragility in the test is to remove >>> >>> > > the >>> >>> > > running of `short.py` in the test. This is only invoked to check >>> >>> > > that >>> >>> > > it's possible for a command to run to completion in the presence of >>> >>> > > a >>> >>> > > fixed timeout. If we can live without testing that part (i.e. we >>> >>> > > only >>> >>> > > test that a timeout can be reached) then the test should be much >>> >>> > > more >>> >>> > > robust. >>> >>> > >>> >>> > If you're on board with that, it's a tradeoff I think is probably >>> >>> > reasonable from a test coverage V reliability V development time >>> >>> > tradeoff. >>> >>> >>> >>> Sorry for the delay here. I've put a patch up for review that goes >>> >>> with this approach: https://reviews.llvm.org/D88807 >>> >> >>> >> ___ >>> >> LLVM Developers mailing list >>> >> llvm-...@lists.llvm.org >>> >> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev >>> ___ >>> LLVM Developers mailing list >>> llvm-...@lists.llvm.org >>> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev ___ linaro-toolchain mailing list linaro-toolchain@lists.linaro.org https://lists.linaro.org/mailman/listinfo/linaro-toolchain
Re: Timeout tests timing out
> > One thing we could do to remove fragility in the test is to remove the > > running of `short.py` in the test. This is only invoked to check that > > it's possible for a command to run to completion in the presence of a > > fixed timeout. If we can live without testing that part (i.e. we only > > test that a timeout can be reached) then the test should be much more > > robust. > > If you're on board with that, it's a tradeoff I think is probably > reasonable from a test coverage V reliability V development time > tradeoff. Sorry for the delay here. I've put a patch up for review that goes with this approach: https://reviews.llvm.org/D88807 ___ linaro-toolchain mailing list linaro-toolchain@lists.linaro.org https://lists.linaro.org/mailman/listinfo/linaro-toolchain
Re: [llvm-dev] Timeout tests timing out
On Wed, 16 Sep 2020 at 22:32, Mehdi AMINI wrote: > > Hi, > > Depending on timing in a test is quite brittle in general: can we mock the > timeout instead and make this fully deterministic somehow? > > The test has this comment: > > > FIXME: This test is fragile because it relies on time which can > > be affected by system performance. In particular we are currently > > assuming that `short.py` can be successfully executed within 2 > > seconds of wallclock time. > > Maybe "short.py" can be replaced by adding into lit itself a "no op" which > would just not really spawn a process and instead mark the task as completed > immediately internally? I think it's probably easier to just remove running "short.py" in the test. We'd lose some test coverage because we no longer test that a program can run to completion when a timeout is set but maybe this is an acceptable trade-off to avoid the test being flakey? ___ linaro-toolchain mailing list linaro-toolchain@lists.linaro.org https://lists.linaro.org/mailman/listinfo/linaro-toolchain
Re: Timeout tests timing out
On Wed, 16 Sep 2020 at 22:24, David Blaikie wrote: > > I appreciate the value of the feature - but it's possible the test > doesn't pull its weight. Is the code that implements the feature > liable to failure/often touched? If it's pretty static/failure is > unlikely, possibly the time and flaky failures aren't worth the value > of possibly catching a low-chance bug. I don't have many data points here. I haven't needed to touch the code or tests since 2018. I haven't been asked to review any patches related to the timeout feature since then so I presume nobody has touched it since. > Another option might be to reduce how often/in which configurations > the test is run - LLVM_ENABLE_EXPENSIVE_CHECKS presumably only works > for code within LLVM itself, and not test cases - but maybe I'm wrong > there & this parameter could be used (& then the timing bumped up > quite a bit to try to make it much more reliable), or something > similar could be implemented at the lit check level? > > Ah, compiler-rt tests use EXPENSIVE_CHECKS to disable certain tests: > > ./compiler-rt/test/lit.common.configured.in:set_default("expensive_checks", > @LLVM_ENABLE_EXPENSIVE_CHECKS_PYBOOL@) > ./compiler-rt/test/fuzzer/large.test:UNSUPPORTED: expensive_checks > > Could you bump the timeouts a fair bit and disable the tests except > under expensive checks? Only running that test in the "expensive_checks" configuration kind of abuses that configuration option because the test is actually not an expensive check. It's just that the test is sensitive to the performance of the running system. If the system is under heavy load when the test is running it's more likely to fail. One thing we could do to remove fragility in the test is to remove the running of `short.py` in the test. This is only invoked to check that it's possible for a command to run to completion in the presence of a fixed timeout. If we can live without testing that part (i.e. we only test that a timeout can be reached) then the test should be much more robust. If for some reason that's not enough another thing that might help with this problem is to not run the lit tests at the same time as all the other tests. The hope here is that the lit tests would run while the system load is lower. We could do this by changing the build system to * Make all `check-*` (except `check-lit`) targets depend on the `check-lit` target. This would ensure that lit gets tested before we test everything else in LLVM. * Remove the lit test directory from inclusion in all `check-*` targets (except `check-lit`) so we don't test lit twice. This also has a nice benefit of testing lit **before** we use it to test everything else in LLVM. Today we don't do that, we just run all the tests in one big lit invocation (i.e. we test lit at the same time that we are using it). It does have some downsides though * If any of the lit tests fail we won't run the rest of the tests so you won't get the results of running the other tests until the lit tests are fixed. * When running any of the `check-*` targets (apart from `check-lit`) you have to wait for the lit tests to run before any other tests run. What do you think? > On Wed, Sep 16, 2020 at 9:31 PM Dan Liew wrote: > > > > Hi David, > > > > Unfortunately writing a reliable test is tricky given that the > > functionality we're trying to test involves timing. I would advise > > against disabling the test entirely because it actually tests > > functionality that people use. I'd suggest bumping up the time limits. > > This is what I've done in the past. See > > > > commit 6dfcc78364fa3e8104d6e6634733863eb0bf4be8 > > Author: Dan Liew > > Date: Tue May 22 15:06:29 2018 + > > > > [lit] Try to make `shtest-timeout.py` test more reliable by using a > > larger timeout value. This really isn't very good because it will > > still be susceptible to machine performance. > > > > While we are here also fix a bug in validation of > > `maxIndividualTestTime` where previously it wasn't checked if the > > type was an int. > > > > rdar://problem/40221572 > > > > llvm-svn: 332987 > > > > HTH, > > Dan. > > > > On Wed, 16 Sep 2020 at 09:37, David Blaikie wrote: > > > > > > Ping on this > > > > > > On Wed, Sep 9, 2020 at 8:27 PM David Blaikie wrote: > > > > > > > > The clang-cmake-armv8-lld (linaro-toolchain owners) buildbot is timing > > > > out trying to run some timeout tests (Dan Liew author): > > > > > > > > Pass: > > > > http://lab.llvm.org:8011/builders/clang-cmake-armv8-lld/builds/5672 > > > > Fail: > > > > http://lab.llvm.org:8011/builders/clang-cmake-armv8-lld/builds/5673 > > > > > > > > Is there anything we can do to the buildbot? Or the tests? (bump up the > > > > time limits or maybe remove the tests as unreliable?) ___ linaro-toolchain mailing list linaro-toolchain@lists.linaro.org https://lists.linaro.org/mailman/listinfo/linaro-toolchain
Re: Timeout tests timing out
Hi David, Unfortunately writing a reliable test is tricky given that the functionality we're trying to test involves timing. I would advise against disabling the test entirely because it actually tests functionality that people use. I'd suggest bumping up the time limits. This is what I've done in the past. See commit 6dfcc78364fa3e8104d6e6634733863eb0bf4be8 Author: Dan Liew Date: Tue May 22 15:06:29 2018 + [lit] Try to make `shtest-timeout.py` test more reliable by using a larger timeout value. This really isn't very good because it will still be susceptible to machine performance. While we are here also fix a bug in validation of `maxIndividualTestTime` where previously it wasn't checked if the type was an int. rdar://problem/40221572 llvm-svn: 332987 HTH, Dan. On Wed, 16 Sep 2020 at 09:37, David Blaikie wrote: > > Ping on this > > On Wed, Sep 9, 2020 at 8:27 PM David Blaikie wrote: > > > > The clang-cmake-armv8-lld (linaro-toolchain owners) buildbot is timing out > > trying to run some timeout tests (Dan Liew author): > > > > Pass: http://lab.llvm.org:8011/builders/clang-cmake-armv8-lld/builds/5672 > > Fail: http://lab.llvm.org:8011/builders/clang-cmake-armv8-lld/builds/5673 > > > > Is there anything we can do to the buildbot? Or the tests? (bump up the > > time limits or maybe remove the tests as unreliable?) ___ linaro-toolchain mailing list linaro-toolchain@lists.linaro.org https://lists.linaro.org/mailman/listinfo/linaro-toolchain