The weston runtime test is unreliable, depending on the execution speed
of its runner. Example of failure is [1], and success is [2]. This
commit improves the test in several ways, to make it more robust to
execution speed variations:
- The command started in background (weston, weston-simple-egl) are
now started in a subshell. This suppresses the job control messages
when they are stopped. Those messages could interfere with the parsing
of the output;
- Wait time are moved outside the emulator;
- The kernel argument vt.global_cursor_default=0 is added, to make sure
cursors are globally disabled, since the test use vkms display CRCs;
- The memory of the emulator is increased to 512M. This test uses cpio
initramfs, and the filesystem size increased;
- The vkms driver emulates a "vsync" event, but can generate a warning
when the system is too slow. This warning is printed on the console by
the klogd daemon in its default configuration. This commit adds the
overlay file /etc/default/klogd to limit only kernel emergency
messages to be printed on the console. This change fixes the failure
seen in [1] ;
- Some sleep time values were adjusted to run on a "performant" idle
host. Those values are not suitable in all situations. In the
meantime a generic retry mechanism is added in the test infra, this
commit adds a retry logic inspired from the test_flutter runtime
test.
Fixes: [1]
[1] https://gitlab.com/buildroot.org/buildroot/-/jobs/8562483474
[2] https://gitlab.com/buildroot.org/buildroot/-/jobs/8435236652
Signed-off-by: Julien Olivain <ju.o@free.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
The lmbench package (largely unmaintained upstream, last release in
2006) uses a home-grown build system. The scripts/build script sets
LDLIBS to -lm and Buildroot's makefile changes that to:
LDLIBS+=-lm
Except that when Buildroot passes LDLIBS="`$(PKG_CONFIG_HOST_BINARY)
--libs libtirpc`", the LDLIBS variable ends up being equal to:
-ltirpc-lm
Yes, without any space between -ltirpc and -lm.
Due to this, the checks in scripts/build that use ${LDLIBS} all fail,
and in particular the test that checks when socklen_t is a known type
fails, causing lmbench to provide its own definition, which clashes
with the C library headers definition, and therefore causing build
failures such as:
bench.h:81:13: error: conflicting types for 'socklen_t'; have 'int'
This commit fixes that by adjusting scripts/build using a patch to
properly allow passing additional ${LDLIBS} value, with the needed
space to separate the value from -lm.
Fixes:
http://autobuild.buildroot.net/results/f1715de95b46a1d08143e529bd4574bc7dbcfb3e/
We have been unable to determine exactly when this issue was
introduced. The first build failure we could find is the one
referenced above, which dates back from Aug 19, 2024. Since this date,
lmbench has been consistently failing on a very regular
basis. However, prior to Aug 19, 2024, the previous failure was from
December 2022, and was unrelated. It is unclear what changed in Aug
2024 to cause this issue to surface. The one thing that changed right
before commit ce3dedc2 (first failing commit) are changes to
genrandconfig, which ensures all autobuilders now generated fully
random configurations instead of configurations based on a
well-defined list of arch/toolchain configurations. But even with
this, this lmbench issue should have appeared earlier, and we have
been unable to find a scientific explanation.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
The android-tools code base uses the __b64_pton() function, which
isn't provided by all C libraries. So the Debian patch
debian/patches/add_adbd.patch adds an implementation of b64_pton(),
but doesn't actually use it, nor defines a prototype for it. Our
existing patch 0003-Fix-build-issue-with-uclibc.patch switches the
code to use the b64_pton() function... but still without providing a
prototype, causing the following build failures with GCC >= 14.x:
adb_auth_client.c:75:15: error: implicit declaration of function 'b64_pton'
To fix this, we rework 0003-Fix-build-issue-with-uclibc.patch into a
patch that:
(1) Renames b64_pton() to adb_b64_pton() to make sure it won't clash
with implementations provided by some C libraries, and adjusts
the call sites accordingly.
(2) Adds a prototype definition of adb_b64_pton() in places where
this function is used.
Fixes:
http://autobuild.buildroot.net/results/b25b25337c7ad89c33f8bd20b646850bd993ec53ae9/
Even though GCC 14.x support was merged in Buildroot in May 2024, this
particular b64_pton() only started appearing on July 15 2024, with the
first occurence being:
http://autobuild.buildroot.net/results/1cbe87bbe3c56f28444b3aaba1ba1d05f947d36e/
Indeed, it's not before July 15 2024 that we merged commit
d201f2f5cd ("package/android-tools: add
patches to fix build with GCC 14.x"), which fixed other GCC 14.x
issues, which were hiding this b64_pton() problem.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Arnout Vandecappelle <arnout@mind.be>
The kernel headers on mips64 and powerpc64 were historically "broken",
defining u64 as an "unsigned long" instead of "unsigned long
long". This has been fixed in the upstream Linux kernel by introducing
the __SANE_USERSPACE_TYPES__ definition.
Our commit a2e178d6b4 ("android-tools:
disable on some architecture with old kernel headers") already
disabled building fastboot on powerpc64 and mips64 systems that have
too old kernel headers. However, it turns out that even with the new
kernel headers, there are build issues because
__SANE_USERSPACE_TYPES__ isn't defined everywhere it should be in the
android-tools code base. The Debian patche
debian/patches/ppc64el-ftbfs.patch adds some definitions, but it's
missing one file, and in another the definition comes too late.
This commit adds an extra patch that fixes this up, and makes fastboot
build properly on mips64 and powerpc64.
Fixes:
http://autobuild.buildroot.net/results/8528ff876e695f79bdfe64f5330d9d51eeef66cb/ (powerpc64)
http://autobuild.buildroot.net/results/36ac6af73b618c28d1636093da333f7ebd9d6cfe/ (mips64)
This issue has been occurring as far as Feb 2021, with the first
occurence apparently being:
http://autobuild.buildroot.net/results/d9521b4bfeafb1140c21745dbfe28d476a9b71ec/
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Arnout Vandecappelle <arnout@mind.be>
This will make it easier to modify the patch series moving forward. No
functional changes.
On one patch, we added the SoB from Giulio, which was missing. Giulio
was the original contributor of the patch, and nobody touched it since
it was introduced.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Arnout Vandecappelle <arnout@mind.be>
Upstream moved old versions in an "older" directory. The current version
in Buildroot was moved to this directory and fail to download.
This commit simply updates the _SITE url to reflect this change.
Fixes:
https://gitlab.com/buildroot.org/buildroot/-/jobs/8514189435
Signed-off-by: Julien Olivain <ju.o@free.fr>
Signed-off-by: Arnout Vandecappelle <arnout@mind.be>
losetup -f returns the next free loop device, which may not be
/dev/loop0. If you blindly follow the readmy you may end up destroying
an existing device.
Make it more robust with a variable to store the loop device.
Signed-off-by: Cherniaev Andrei <dungeonlords789@naver.com>
[Arnout: keep the actual losetup atomic]
Signed-off-by: Arnout Vandecappelle <arnout@mind.be>
nfs-utils configure script uses "systemd-escape" to determine the
rpc_pipefs mount point unit name in the non-default case.
If the host build system has no systemd this will silently result in an
empty name, causing rpc_pipefs.target to incorrectly depend on ".mount".
Fix that by depending on host-systemd for build.
Signed-off-by: Anssi Hannula <anssi.hannula@bitwise.fi>
Acked-by: Petr Vorel <petr.vorel@gmail.com>
Signed-off-by: Arnout Vandecappelle <arnout@mind.be>
Commit efe93b34e "package/opkg: bump to version 0.7.0" [1] introduced
a use of --without-xattrs (with an extra 's') configure option.
When compiling opkg, the configure step show the warning:
configure: WARNING: unrecognized options: ... --without-xattrs
The actual option name is "--without-xattr" (without an extra 's').
See [2].
This commit fixes the typo.
[1] efe93b34ec
[2] https://git.yoctoproject.org/opkg/tree/configure.ac?h=v0.7.0#n167
Signed-off-by: Julien Olivain <ju.o@free.fr>
Signed-off-by: Arnout Vandecappelle <arnout@mind.be>
When the rust package was added in Buildroot it was using
the 1.23.0 release [1] where some vendored sources contained a
Cargo.toml.orig file that caused issues with
support/scripts/apply-patches.sh used by Buildroot package
infrastructure.
Since then, Buildroot is now using rust 1.74.1 (2024.02.x),
1.79.0 (2024.08.x) and 1.82.0 (master) where vendored
sources no longer contains any Cargo.toml.orig file.
So this post-extract hook is no longer needed.
Moreover, since Rust 1.81.0 are present in
src/tools/rustc-perf/collector/compile-benchmarks/<vendored pkg>
where there is no .cargo-checksum.json file.
Since then, the sed command fail and stop the build:
sed: can't read [...]/host-rust-1.81.0/src/tools/rustc-perf/collector/compile-benchmarks/serde_derive-1.0.136/.cargo-checksum.json: No such file or directory
Fixes:
https://gitlab.com/buildroot.org/buildroot/-/jobs/8232367821
[1] b50522d838
Signed-off-by: Romain Naour <romain.naour@smile.fr>
Signed-off-by: Arnout Vandecappelle <arnout@mind.be>
The particular combination of LVM2 (which enables -PIE), musl-libc with
dynamic linking (and thus -ztext, banning TEXTRELs), and microblaze
leads to an error because a TEXTREL is produced in musl-libc's Scrt1.o.
The upstream patch added in this commit fixes (or works around) the
problem.
Fixes: http://autobuild.buildroot.net/results/5bb283eff4b4a2b17d1028bbe0b3e1bea8fbeba8/
Signed-off-by: J. Neuschäfer <j.neuschaefer@gmx.net>
Signed-off-by: Arnout Vandecappelle <arnout@mind.be>
As it happens, I am unable to keep maintaining those entries, and it is
misleading to others about whether they should Cc me on their changes.
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Signed-off-by: Julien Olivain <ju.o@free.fr>
