| Message ID | a04f80a248fb1452841a424e850a694419e3a6dd.1671873478.git.yann.morin.1998@free.fr |
|---|---|
| State | Accepted |
| Headers | show |
| Series | support/testing: misc improvements (branch yem/runtime-test-ppd) | expand |
On Sat, 24 Dec 2022 10:18:12 +0100 "Yann E. MORIN" <yann.morin.1998@free.fr> wrote: > Running tests with top-level parallel builds can speed up running some > tests, expecially those that have a lot of packages like the systemd > init tests. > > Trigger TLPB when the configuration enables per-package directories. > > Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr> > Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com> I certainly support the idea of running some tests with TLPB. However, this implementation makes a confusion between two different settings: - The existing BRConfigTest.jlevel, which is set by run-tests -j, and used to define BR2_JLEVEL in the Buildroot configuration of each test case. This determines the number of parallel jobs used to build each independent package. - Your new work, which uses "make -j" to do TLPB... but relies on the same above value, even though it's a completely different setting. Is this expected? Best regards, Thomas
Thomas, All, On 2022-12-27 21:45 +0100, Thomas Petazzoni via buildroot spake thusly: > On Sat, 24 Dec 2022 10:18:12 +0100 > "Yann E. MORIN" <yann.morin.1998@free.fr> wrote: > > Running tests with top-level parallel builds can speed up running some > > tests, expecially those that have a lot of packages like the systemd > > init tests. > > > > Trigger TLPB when the configuration enables per-package directories. > > > > Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr> > > Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com> > > I certainly support the idea of running some tests with TLPB. However, > this implementation makes a confusion between two different settings: > > - The existing BRConfigTest.jlevel, which is set by run-tests -j, and > used to define BR2_JLEVEL in the Buildroot configuration of each test > case. This determines the number of parallel jobs used to build each > independent package. > > - Your new work, which uses "make -j" to do TLPB... but relies on the > same above value, even though it's a completely different setting. > > Is this expected? Yes, this is the intended behaviour, which I was explicitly seeking. So, if one runs with PPD and TLPB (outside the run-time infra), one would do something like: $ make -jN This spawns a top-level make process that is parallel. In turn, in rules, when we call $(MAKE), this is the magic that tells make that it is recursive, but that it should use the jobserver from the calling process. So, in this case, the BR2_JLEVEL is unused by whatever uses the make jobserver; only the number of jobs in the top-level jobserver is meaningful, i.e. whatever we pass as -jN. The exception, of course, is whatever uses BR2_JLEVEL but does not talk to the jobserver, but this is mostly a few packages (scons, waf et al.). Even ninja packages do talk to the top-level jobserver, now that we use the ninja fork that knows to talk to it. So, yes, using top-level -jN with the same value as BR2_JLEVEL is exactly what I intended to do. Unless I totally missed something... Regards, Yann E. MORIN.
Thomas, All, On 2022-12-27 21:53 +0100, Yann E. MORIN spake thusly: > On 2022-12-27 21:45 +0100, Thomas Petazzoni via buildroot spake thusly: > > On Sat, 24 Dec 2022 10:18:12 +0100 > > "Yann E. MORIN" <yann.morin.1998@free.fr> wrote: > > > Running tests with top-level parallel builds can speed up running some > > > tests, expecially those that have a lot of packages like the systemd > > > init tests. > > > > > > Trigger TLPB when the configuration enables per-package directories. > > > > > > Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr> > > > Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com> > > > > I certainly support the idea of running some tests with TLPB. However, > > this implementation makes a confusion between two different settings: > > > > - The existing BRConfigTest.jlevel, which is set by run-tests -j, and > > used to define BR2_JLEVEL in the Buildroot configuration of each test > > case. This determines the number of parallel jobs used to build each > > independent package. > > > > - Your new work, which uses "make -j" to do TLPB... but relies on the > > same above value, even though it's a completely different setting. > > > > Is this expected? > > Yes, this is the intended behaviour, which I was explicitly seeking. > > So, if one runs with PPD and TLPB (outside the run-time infra), one > would do something like: > > $ make -jN > > This spawns a top-level make process that is parallel. In turn, in > rules, when we call $(MAKE), this is the magic that tells make that > it is recursive, but that it should use the jobserver from the calling > process. > > So, in this case, the BR2_JLEVEL is unused by whatever uses the make > jobserver; only the number of jobs in the top-level jobserver is > meaningful, i.e. whatever we pass as -jN. Slight correction: BR2_JLEVEL does have an actual effect, but only if its value is lower than the one we pass as -jN. So, in practice, we want to have BR2_JLEVEL equal to the top-level -jN. Regards, Yann E. MORIN. > The exception, of course, is whatever uses BR2_JLEVEL but does not talk > to the jobserver, but this is mostly a few packages (scons, waf et al.). > Even ninja packages do talk to the top-level jobserver, now that we use > the ninja fork that knows to talk to it. > > So, yes, using top-level -jN with the same value as BR2_JLEVEL is > exactly what I intended to do. > > Unless I totally missed something... > > Regards, > Yann E. MORIN. > > -- > .-----------------.--------------------.------------------.--------------------. > | Yann E. MORIN | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: | > | +33 662 376 056 | Software Designer | \ / CAMPAIGN | ___ | > | +33 561 099 427 `------------.-------: X AGAINST | \e/ There is no | > | http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL | v conspiracy. | > '------------------------------^-------^------------------^--------------------' > _______________________________________________ > buildroot mailing list > buildroot@buildroot.org > https://lists.buildroot.org/mailman/listinfo/buildroot
On Sat, 24 Dec 2022 10:18:12 +0100 "Yann E. MORIN" <yann.morin.1998@free.fr> wrote: > Running tests with top-level parallel builds can speed up running some > tests, expecially those that have a lot of packages like the systemd > init tests. > > Trigger TLPB when the configuration enables per-package directories. > > Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr> > Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com> > --- > support/testing/infra/basetest.py | 2 +- > support/testing/infra/builder.py | 5 ++++- > 2 files changed, 5 insertions(+), 2 deletions(-) Thanks to the additional explanation you provided, I understood better the idea, and therefore applied the patch. I updated the commit log a little bit to summarize the explanation you gave, and applied. Thanks! Thomas
diff --git a/support/testing/infra/basetest.py b/support/testing/infra/basetest.py index 45bcd4c2e2..670c7213d6 100644 --- a/support/testing/infra/basetest.py +++ b/support/testing/infra/basetest.py @@ -52,7 +52,7 @@ class BRConfigTest(unittest.TestCase): def setUp(self): self.show_msg("Starting") - self.b = Builder(self.config, self.builddir, self.logtofile) + self.b = Builder(self.config, self.builddir, self.logtofile, self.jlevel) if not self.keepbuilds: self.b.delete() diff --git a/support/testing/infra/builder.py b/support/testing/infra/builder.py index 922a707220..a2abb9ed89 100644 --- a/support/testing/infra/builder.py +++ b/support/testing/infra/builder.py @@ -6,11 +6,12 @@ import infra class Builder(object): - def __init__(self, config, builddir, logtofile): + def __init__(self, config, builddir, logtofile, jlevel=None): self.config = '\n'.join([line.lstrip() for line in config.splitlines()]) + '\n' self.builddir = builddir self.logfile = infra.open_log_file(builddir, "build", logtofile) + self.jlevel = jlevel def is_defconfig_valid(self, configfile, defconfig): """Check if the .config is contains all lines present in the defconfig.""" @@ -87,6 +88,8 @@ class Builder(object): env.update(make_extra_env) cmd = ["make", "-C", self.builddir] + if "BR2_PER_PACKAGE_DIRECTORIES=y" in self.config.splitlines() and self.jlevel: + cmd.append(f"-j{self.jlevel}") cmd += make_extra_opts ret = subprocess.call(cmd, stdout=self.logfile, stderr=self.logfile,
Running tests with top-level parallel builds can speed up running some tests, expecially those that have a lot of packages like the systemd init tests. Trigger TLPB when the configuration enables per-package directories. Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr> Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com> --- support/testing/infra/basetest.py | 2 +- support/testing/infra/builder.py | 5 ++++- 2 files changed, 5 insertions(+), 2 deletions(-)