Use ccache to speedup kernel compilation.
Ccache dir is cached across each build test. To refresh ccache directory
we generate an hash of the kernel include files, that includes the
kernel versions of every kernel supported and the kernel compile
includes.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
(cherry picked from commit 137ba15e6e)
Enable ccache support for tools container, useful to speedup other
workflow even more.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
(cherry picked from commit 2781e3839e)
Use tools:latest container with prebuilt host tools to speedup kernel
compilation in kernel workflow.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
(cherry picked from commit 5d09118f8e)
Currently each Kernel compilation takes about 30 minutes of which 20
minutes are used to compile our tools. While the toolchain is downloaded
and instantly ready the tools are missing.
This commit starts uploading a Docker container including compiled tools
which are ready to use. It is automatically updated whenever any tools
are changed.
Signed-off-by: Paul Spooren <mail@aparcar.org>
Co-Developed-by: Christian Marangi <ansuelsmth@gmail.com>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
(cherry picked from commit 25b65f548d)
Instead of using a fresh Linux installation which is setup every time
use the Buildbot container which is used for our own Buildbot
infrastructure, too.
While at it also tidy up the workflow to make it more consistent with
other workflow.
Signed-off-by: Paul Spooren <mail@aparcar.org>
Co-Developed-by: Christian Marangi <ansuelsmth@gmail.com>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
(cherry picked from commit 3b23227d43)
Instead of waiting for complete workflow finish, then downloading the
artifacts, unpacking them and inspecting them, lets try to make the
build failure immediately visible in the log output:
====== Make errors from logs/target/linux/compile.txt ======
* Legacy (non-UHI/non-FIT) Boards
*
Support MIPS SEAD-3 boards (LEGACY_BOARD_SEAD3) [N/y/?] (NEW)
Error in reading or end of file.
make[6]: *** [scripts/kconfig/Makefile:77: syncconfig] Error 1
make[5]: *** [Makefile:616: syncconfig] Error 2
make[4]: *** [Makefile:736: include/config/auto.conf.cmd] Error 2
make[3]: *** [Makefile:24: build_dir/target-mipsel-openwrt-linux-musl_musl/linux-ramips_mt7620/linux-5.15.62/.modules] Error 2
make[2]: *** [Makefile:11: compile] Error 2
time: target/linux/compile#30.09#11.30#37.92
Signed-off-by: Petr Štetiar <ynezz@true.cz>
(cherry picked from commit f4ca4187cd)
Grant pull-requests write permission to the labeler workflow and
read-only to everything else.
Signed-off-by: Alex Low <aleksandrosansan@gmail.com>
[ wrap to 80 columns and fix wrong author as requested by author itself ]
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
(cherry picked from commit 7152599407)
This adds GitHub CI action which makes use of 'Labeler', allowing
automatic labeling of new PRs, based on the modified files paths.
Below labels are supported and more can be added later:
- 'target/*'
- 'target/imagebuilder'
- 'kernel'
- 'core packages'
- 'build/scripts/tools'
- 'toolchain'
- 'GitHub/CI'
For more information:
https://github.com/marketplace/actions/labeler
Signed-off-by: Piotr Dymacz <pepe2k@gmail.com>
(cherry picked from commit 4f42566d47)
Actually package kmods in kernel workflow to catch dependency error and
other problem that may arise from kmods packaging.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
(cherry picked from commit 0c45db5560)
Cache external toolchain for each target to remove load from openwrt cdn
server and make the external toolchain setup quicker.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
(cherry picked from commit f522c27385)
Activate building all kernel modules.
This builds all kernel modules from the core packages and the feeds.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
(cherry picked from commit fed325f463)
Instead of cloning the feeds from the default location at
git.openwrt.org use the github action to clone them directly from
github. We saw some error messages when cloning from git.openwrt.org,
probably related to some rate limiting applied. Cloning from github
within a github action should work more stable.
The "./scripts/feeds update -a" script will use the already checked out
feed repositories and not clone them again from git.openwrt.org, but it
will also not change the branch name.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
(cherry picked from commit b120e78917)
Show the configuration used to build OpenWrt before starting the build.
This should make it easier for people to reproduce problems when it
fails.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
(cherry picked from commit 43afaf6149)
Use downloads.cdn.openwrt.org to download the toolchain. This should
reduce the load on the servers.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
(cherry picked from commit 7b4daf0076)
Run github actions insider buildbot docker container.
Signed-off-by: Paul Spooren <mail@aparcar.org>
[ run container under buildbot user ]
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
(cherry picked from commit 8a77adb048)
Add Github Actions yaml script to build test kernel PR changes for
each target.
Signed-off-by: Paul Spooren <mail@aparcar.org>
[ add commit description ]
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
(cherry picked from commit c17c931a90)
* Always store build logs
* Store .config as an artifact
* Rename job to `tools-{ os }` for log archive without spaces
* Run CI job on changes to the CI file itself
Signed-off-by: Paul Spooren <mail@aparcar.org>
(cherry picked from commit 80f79beb95)
Artifacts can only be uploaded from inside the GITHUB_WORKSPACE. While
the Linux CI jobs run inside that per default, a special case-sensitive
mount outside the GITHUB_WORKSPACE is used for macOS builds.
To make log artifacts work for both macOS and Linux, move logs/ folder
to GITHUB_WORKSPACE on failures.
Signed-off-by: Paul Spooren <mail@aparcar.org>
(cherry picked from commit a5eeac8033)
This commits adds GitHub CI to check that all tools compile on both
Ubuntu and macOS. Since running in parrallel this should also detect
badly set depdendencies within tools/Makefile.
Signed-off-by: Paul Spooren <mail@aparcar.org>
(cherry picked from commit fb830fd894)
Due to a limited grep pattern, subjects containing dots like `image.mk`
were falsely reported as problematic. Extend pattern to allow dots.
Signed-off-by: Paul Spooren <mail@aparcar.org>
(cherry picked from commit fb5d0dc362)
The formal checks verify the following things:
- Commits does not contain any merge commits
- Signed by a real name
- Commit titles starts with an `<area>:`
- Author name matches signed of name
- Commit message is not empty
Signed-off-by: Paul Spooren <mail@aparcar.org>
(cherry picked from commit 3dc38823a2)
Openwrt now supports only glibc and musl. Add support for musl and
rework the libc check to handle the new config flags and correctly
compile package basend on that.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
(cherry picked from commit 7be01fe13b)
Openwrt generate info.mk that contains the libc type. For probe_cc check
if the file exist and parse directly it for LIBC type.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
(cherry picked from commit 75311977f5)
Currently we never call probe_cc before config generation, this cause
the script to never actually detect the correct libc type.
Call probe_cc before config generation to correctl set the .config file.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
(cherry picked from commit ddeabc75eb)
It can be useful to overwrite an already generated config.
Option are simply added at the end of the config and make defconfig
will overwrite the relevant option with the new one.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
(cherry picked from commit f4dd18ca39)
The parsed prefix in print_config is wrong and this produce broken
generated .config that won't work with any external toolchain.
Currently the prefix from a CC of
'arm-openwrt-linux-muslgnueabi-gcc-12.1.0'
produce a prefix
'arm-openwrt-linux-muslgnueabi-gcc-'
This is wrong as the real prefix should be
'arm-openwrt-linux-muslgnueabi-'
This is probably caused by a change in how the toolchain is now handled
that now append also the gcc version. Probably in ancient days the
version wasn't part of the name and the prefix generation stripped the
'-gcc' instead of the gcc version.
Fix this and correctly strip the gcc version and the gcc suffix to
correctly call toolchain bins.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
(cherry picked from commit 53c293262f)
This only affects typos in comments or user-facing output.
Signed-off-by: Josh Soref <jsoref@users.noreply.github.com>
[only picks changes to scripts, drop "commandline" replacement,
fix case for "arbitrary", improve commit message]
Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
(cherry picked from commit 08622de7d6)
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Add a simple script to make it easier to install a prebuilt tools tar.
Currently it will be used by our tools container and kernel workflow on
github.
Simple script that take a tar that contains prebuilt host tools, extract
them and refresh the timestamps to skip recompilation of such host
tools.
By default it refresh timestamps of build_dir/host and
staging_dir/host/stamp.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
(cherry picked from commit 5428bdc2df)
When building with an external toolcahin with musl also include
./include/fortify by default. This is also done when we build with the
internal toolchain using musl libc.
Without this extra include the fortify source feature is not working
when using an external musl toolchain. All binaries were compiled
without fortify source when an external musl toolchain was used. All
binaries release done by the OpenWrt project use the internal toolcahin
where fortify source is working.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
(cherry picked from commit 3053045093a5c89dad970b7db3ed9f2a5be993f6)
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
When using the OpenWrt toolchain as an external toolchain the build
failed due to missing LTO support. By choosing the GCC wrappers of
the tools this commit makes sure that the LTO-enabled executables
are being used.
Signed-off-by: Vincent Wiemann <vincent.wiemann@ironai.com>
[ wrap the commit description to 72 char ]
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
(cherry picked from commit 2555ffb453)
Don't add wrapped bin to the TARGET_PATH as it does cause compilation
error.
cmake.mk will use the "command -v" and will use the wrapped bin instead
of the external toolchain bin as they have the same name and command
will select the first result.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
(cherry picked from commit a90eabf602)
While experimenting with the AUTOREMOVE option in search of a way to use
prebuilt host tools in different buildroot, it was discovered that the
md5 generated by find_md5 in depends.mk is not reproducible.
Currently the hash is generated by the path of the file in addition to
the file mod time. Out of confusion, probably, there was an idea that
such command was used on the package build_dir. Reality is that this
command is run on the package files. (Makefile, patches, src)
This is problematic because the package Makefile (for example) change at
each git clone and base the hash on the Makefile mtime doesn't really
reflect if the Makefile actually changes across a buildroot or not.
A better approach is to generate an hash of each file and then generate
an hash on the sort hash list. This way we remove the problem of git
clone setting a wrong mtime while keeping the integrity of checking if a
file changed for the package as any change will result in a different
hash.
Introduce a new kind of find_md5 function, find_md5_reproducible that
apply this new logic and limit it only with AUTOREMOVE option set to
prevent any kind of slowdown due to additional hash generation.
(cherry picked from commit 53a08e3743)
[ reworked mkhash to old implementation ]
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
procd complain for an unused verbose variable causing compilation error.
Fix this by setting the variable static following upstream procd
changes.
This is a variant of 0ee73b2c86a853ae3274c7080e2dcd36b81aa1fa that
introduced major change and fixed the verbose variable to static.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Add missing define for dwc quirk patch required to fix compilation error
for layerscape target. This was wrongly dropped in a kernel bump.
Fixes: 8e5de89769 ("kernel: bump 5.4 to 5.4.213")
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
This fixes:
[ 10.440495] bcm_ns_usb2 1800c000.usb2-phy: can't request region for resource [mem 0x1800c000-0x1800cfff]
[ 10.450039] bcm_ns_usb2 1800c000.usb2-phy: Failed to map DMU regs
[ 10.456183] bcm_ns_usb2: probe of 1800c000.usb2-phy failed with error -16
caused by conflict in allocating resources.
Fixes: 44ce70f0e2 ("bcm53xx: drop downstream patch that now breaks pinctrl driver")
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
(cherry picked from commit 1d7c1a4198)
This includes:
1. BCM5301X changes from 5.14 and queued 5.15 stuff
2. NSP changes from 5.11 - 5.15 for kernel 5.10
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
(cherry picked from commit bce00f47e5)
The most affecting change is move of files from bcm4908/ to the bcmbca/.
That required updating few paths.
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
(cherry picked from commit 45ac906c64)
This helps choosing the right NTFS driver from two available options.
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
(cherry picked from commit b066ad7d9a)
It allows prepopulating /etc/config/network interface-s with predefined
metric. It may be useful for devices with multiple WAN ports.
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
(cherry picked from commit 7f443d2d9a)
