| Message ID | 20200910103725.1439-1-luoyonggang@gmail.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | W32, W64 msys2/mingw patches | expand |
On 10/09/2020 12.37, Yonggang Luo wrote: > This reverts commit 45f7b7b9f38f5c4d1529a37c93dedfc26a231bba > ("cirrus.yml: Split FreeBSD job into two parts"). > > freebsd 1 hour limit not hit anymore > > I think we going to a wrong direction, I think there is some tests a stall the test runner, > please look at > https://cirrus-ci.com/task/5110577531977728 > When its running properly, the consumed time are little, but when tests running too long, > look at the cpu usage, the cpu usage are nearly zero. doesn't consuming time. > > And look at > https://cirrus-ci.com/task/6119341601062912 > > If the tests running properly, the time consuming are little > We should not hide the error by split them Ok, but before we merge this patch, I'd like to understand (and fix if necessary) what is/was causing the slowdowns. Otherwise we'll continue to see failing CI runs, which is very annoying. Thomas
On Thu, Sep 10, 2020 at 10:18 PM Thomas Huth <thuth@redhat.com> wrote: > On 10/09/2020 12.37, Yonggang Luo wrote: > > This reverts commit 45f7b7b9f38f5c4d1529a37c93dedfc26a231bba > > ("cirrus.yml: Split FreeBSD job into two parts"). > > > > freebsd 1 hour limit not hit anymore > > > > I think we going to a wrong direction, I think there is some tests a > stall the test runner, > > please look at > > https://cirrus-ci.com/task/5110577531977728 > > When its running properly, the consumed time are little, but when tests > running too long, > > look at the cpu usage, the cpu usage are nearly zero. doesn't consuming > time. > > > > And look at > > https://cirrus-ci.com/task/6119341601062912 > > > > If the tests running properly, the time consuming are little > > We should not hide the error by split them > > Ok, but before we merge this patch, I'd like to understand (and fix if > necessary) what is/was causing the slowdowns. Otherwise we'll continue > to see failing CI runs, which is very annoying. > > Then I need to apply your idea first, but I found make check are rather slow without the make -j8 check. And even make V=1 check would be slightly slower than make check. > Thomas > >
On Thu, Sep 10, 2020 at 04:18:10PM +0200, Thomas Huth wrote: > On 10/09/2020 12.37, Yonggang Luo wrote: > > This reverts commit 45f7b7b9f38f5c4d1529a37c93dedfc26a231bba > > ("cirrus.yml: Split FreeBSD job into two parts"). > > > > freebsd 1 hour limit not hit anymore > > > > I think we going to a wrong direction, I think there is some tests a stall the test runner, > > please look at > > https://cirrus-ci.com/task/5110577531977728 > > When its running properly, the consumed time are little, but when tests running too long, > > look at the cpu usage, the cpu usage are nearly zero. doesn't consuming time. > > > > And look at > > https://cirrus-ci.com/task/6119341601062912 > > > > If the tests running properly, the time consuming are little > > We should not hide the error by split them > > Ok, but before we merge this patch, I'd like to understand (and fix if > necessary) what is/was causing the slowdowns. Otherwise we'll continue > to see failing CI runs, which is very annoying. I think we need the test harness to print out the time duration for each test in order to stand a chance of find the slow one. A hack like this could be sufficient: diff --git a/scripts/mtest2make.py b/scripts/mtest2make.py index 9cbb2e374d..9103ae65b9 100644 --- a/scripts/mtest2make.py +++ b/scripts/mtest2make.py @@ -20,7 +20,7 @@ print(''' SPEED = quick # $1 = environment, $2 = test command, $3 = test name, $4 = dir -.test-human-tap = $1 $(if $4,(cd $4 && $2),$2) < /dev/null | ./scripts/tap-driver.pl --test-name="$3" $(if $(V),,--show-failures-only) +.test-human-tap = export then=`date +%s` ; $1 $(if $4,(cd $4 && $2),$2) < /dev/null | ./scripts/tap-driver.pl --test-name="$3" $(if $(V),,--show-failures-only) ; export now=`date +%s` ; delta=`expr $$now - $$then` ; $(if $(V),echo "TIME $$delta seconds",true) .test-human-exitcode = $1 $(PYTHON) scripts/test-driver.py $(if $4,-C$4) $(if $(V),--verbose) -- $2 < /dev/null .test-tap-tap = $1 $(if $4,(cd $4 && $2),$2) < /dev/null | sed "s/^[a-z][a-z]* [0-9]*/& $3/" || true .test-tap-exitcode = printf "%s\\n" 1..1 "`$1 $(if $4,(cd $4 && $2),$2) < /dev/null > /dev/null || echo "not "`ok 1 $3" not sure if there is a nicer way todo this but itworks ok for make check-unit at least Regards, Daniel
diff --git a/.cirrus.yml b/.cirrus.yml index 3dd9fcff7f..690c6882e8 100644 --- a/.cirrus.yml +++ b/.cirrus.yml @@ -1,38 +1,19 @@ env: CIRRUS_CLONE_DEPTH: 1 -freebsd_1st_task: +freebsd_12_task: freebsd_instance: image_family: freebsd-12-1 - cpu: 4 - memory: 4G - install_script: ASSUME_ALWAYS_YES=yes pkg bootstrap -f ; pkg install -y - bash curl cyrus-sasl git glib gmake gnutls gsed - nettle perl5 pixman pkgconf png usbredir + cpu: 8 + memory: 8G + install_script: + - ASSUME_ALWAYS_YES=yes pkg bootstrap -f ; + - pkg install -y bash curl cyrus-sasl git glib gmake gnutls gsed + nettle perl5 pixman pkgconf png usbredir script: - mkdir build - cd build - - ../configure --disable-user --target-list-exclude='alpha-softmmu - ppc64-softmmu ppc-softmmu riscv32-softmmu riscv64-softmmu s390x-softmmu - sparc64-softmmu sparc-softmmu x86_64-softmmu i386-softmmu' - --enable-werror || { cat config.log; exit 1; } - - gmake -j$(sysctl -n hw.ncpu) - - gmake -j$(sysctl -n hw.ncpu) check - -freebsd_2nd_task: - freebsd_instance: - image_family: freebsd-12-1 - cpu: 4 - memory: 4G - install_script: ASSUME_ALWAYS_YES=yes pkg bootstrap -f ; pkg install -y - bash curl cyrus-sasl git glib gmake gnutls gtk3 gsed libepoxy mesa-libs - nettle perl5 pixman pkgconf png SDL2 usbredir - script: - - ./configure --enable-werror --target-list='alpha-softmmu ppc64-softmmu - ppc-softmmu riscv32-softmmu riscv64-softmmu s390x-softmmu - sparc64-softmmu sparc-softmmu x86_64-softmmu i386-softmmu - sparc-bsd-user sparc64-bsd-user x86_64-bsd-user i386-bsd-user' - || { cat config.log; exit 1; } + - ../configure --enable-werror || { cat config.log; exit 1; } - gmake -j$(sysctl -n hw.ncpu) - gmake -j$(sysctl -n hw.ncpu) check
This reverts commit 45f7b7b9f38f5c4d1529a37c93dedfc26a231bba ("cirrus.yml: Split FreeBSD job into two parts"). freebsd 1 hour limit not hit anymore I think we going to a wrong direction, I think there is some tests a stall the test runner, please look at https://cirrus-ci.com/task/5110577531977728 When its running properly, the consumed time are little, but when tests running too long, look at the cpu usage, the cpu usage are nearly zero. doesn't consuming time. And look at https://cirrus-ci.com/task/6119341601062912 If the tests running properly, the time consuming are little We should not hide the error by split them Signed-off-by: Yonggang Luo <luoyonggang@gmail.com> Reviewed-by: Daniel P. Berrangé <berrange@redhat.com> Reviewed-by: Ed Maste <emaste@FreeBSD.org> --- .cirrus.yml | 35 ++++++++--------------------------- 1 file changed, 8 insertions(+), 27 deletions(-)