Linaro provides a number of pre-built GCC toolchains for both 32 and
64bit ARM. Switch to their 2017.02 release of gcc-6.3.1 for both.
Cc: Koen Kooi <koen.kooi@linaro.org>
Signed-off-by: Tom Rini <trini@konsulko.com>
- The catch-all i.MX6 job has been exceeding the time limit again so
split this up further. We now have an i.MX6 job and an
everything-else job.
- The logic we use to say "Freescale and AArch64" can be more clearly
expressed with '&' rather than excluding various other things, so
clear that up.
Signed-off-by: Tom Rini <trini@konsulko.com>
For a long while dtc has warned about various constructs. This is now
leading to log file size being exceeded in travis, and as the majority
of these errors need to be fixed in the kernel, switch to using the
stock device-tree-compiler package.
Signed-off-by: Tom Rini <trini@konsulko.com>
As part of 1905c8fc71 we introduced failures depending on if swig and
libpython-dev are installed or not. To provide coverage for this are of
code in the future ensure we have these packages installed.
Signed-off-by: Tom Rini <trini@konsulko.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
First, there are a number of features in newer QEMU that will allow us
to test a wider range of platforms, so we want to use at least v2.8.0.
Second, making use of a PPA for QEMU fails from time to time. So we
change to checking out and building a copy of QEMU when we know that we
are going to use test.py and need QEMU to be installed. This adds
around 4 minutes per test.py job that we run.
Signed-off-by: Tom Rini <trini@konsulko.com>
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Use embded option because of qemu
Use my repo till Stephen merge it.
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Its easier to watch the output of the build process when the platforms
specific boards are grouped in a separate job. This patch adds a job
for all mvebu boards (arm and aarch64).
Signed-off-by: Stefan Roese <sr@denx.de>
Cc: Tom Rini <trini@konsulko.com>
Reviewed-by: Tom Rini <trini@konsulko.com>
In order to avoid running into the time limit, split the 32bit and 64bit
Freescale boards into separate jobs. We could either pass
"freescale & armv8" to buildman or exclude all of the 32bit CPUs. While
the former is shorter I fear the amount of possible escaping required
would make things less readable.
Signed-off-by: Tom Rini <trini@konsulko.com>
The catch-all job is failing due to time limits depending on factors out
of our control, so move Samsung and Rockchip boards into their own jobs
and then exclude them from the general ARM and AArch64 jobs.
Signed-off-by: Tom Rini <trini@konsulko.com>
We have all the building blocks now to run arbitrary efi applications
in travis. The most important one out there is grub2, so let's add
a simple test to verify that grub2 still comes up.
Signed-off-by: Alexander Graf <agraf@suse.de>
Most of the time when running the sleep test in Travis for
the integratorcp_cm926ejs target I get errors like this:
E assert 2.999901056289673 >= 3
The deviation is tiny, but fails the overall build result. Since
the sleep test is not terribly important as gate keeper for travis
tests, let's just exclude it for this board.
Signed-off-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Tom Rini <trini@konsulko.com>
When running in travis-ci, we want to pass environment configuration to
the tests. These reside in a path available through PYTHONPATH, so let's
define that one to point to the unit test repo.
Signed-off-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Tom Rini <trini@konsulko.com>
Some travis QEMU tests can transfer files between the build directory
and the guest U-Boot instance. For that to work, both need to have access
to the same directory.
This patch puts the current build path into an environment variable, so
that the environment generating python scripts can extract it from there
and read the respective files.
Signed-off-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Tom Rini <trini@konsulko.com>
The way that we have things broken down currently allows for some
combinations of vendor or CPU to not be built. To fix this, create a
new catch-all job that excludes everything we've built elsewhere. For
the sake of simplicity we are allowing for the possibility of some
overlap between the vendor-based jobs and the CPU-based jobs. While
we're in here, make a failed build provide the summary of failure.
Signed-off-by: Tom Rini <trini@konsulko.com>
- Add in system aarch64-linux-gnu toolchain
- Now that all VMs will have aarch64 available, don't exclude them from
other jobs but instead exclude them from the catch-all aarch64 build
- Add JOB= to the Freescale/ARM build to be clear about what it does.
- Add uniphier as a stand-alone job
Signed-off-by: Tom Rini <trini@konsulko.com>
ARMv7 Tegra boards aren't currently covered by any other travis-ci jobs.
Add a new job to build them.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Tom Rini <trini@konsulko.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
Use buildman to compile any U-Boot binary tested by test/py. This
re-uses all the work done elsewhere to make buildman work within
Travis-CI, in particular related to toolchain downloading and buildman
config file creation.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
Invoking exit prevents any subsequent build commands from running, and
future patches will add extra commands.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
This places build results into a board-specific directory rather than a
buildman-thread-specific directory. This is required so that we can
access the directory from test.py, and there's no risk of a particular
build's results being over-written by another build performed by the
same thread.
In theory, this can lead to slower builds when building many different
boards in a single buildman thread, since it removes the possibility of
incremental builds between boards. In practice however I didn't notice
longer build times when when enabling this option; if anything build
times decreased although I suspect that's simply due to general
variations in build performance across different machines within the
Travis CI infra-structure.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
Any time an x86 toolchain is used, we need to edit ~/.buildman to
reference it. Move the editing logic into a central place so that it
doesn't have to be duplicated everywhere that uses the x86 toolchain;
future patches will add additional cases where it's used.
It would be nice if we could unconditionally write all of ~/.buildman at
once. Unfortunately, buildman fails if any toolchain mentioned in a
toolchain-prefix entry doesn't exist, even if it doesn't need to use it
for the current build.
The sandbox/x86 build definition currently does nothing more than edit
~/.buildman; no builds are run. Fix this by not defining a custom script
for this build, and hence preventing that stanza from replacing the
default script.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
The phrase "if [ $? -ne 0 ]; then exit $?; fi" doesn't work correctly;
by the time the "exit" statement runs, $? has already been over-written
by the result of the [ command. Fix this by explicitly storing $? and
then using that stored value in both the test and the error-case exit
statement.
This change also converts from textual comparison to integer comparison,
since the exit code is an integer and there's no need to convert it to
a string for comparison.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
Travis CI seems to be confused when there's a colon in an echo command,
and this is currently worked around using a variable that contains the
text we want to echo. Use = syntax instead so that we can remove the
work-around; it's rather confusing until you find out what it's for.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
There were two sub-jobs to build arm1136. Remove the duplicate.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Tom Rini <trini@konsulko.com>
Travis CI names sub-jobs after the first environment variable that is set
for a script. This doesn't produce meaningful results for any of the non-
buildman jobs. Add a dummy variable to give the jobs meaningful names.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Tom Rini <trini@konsulko.com>
- Add a PPA for a more recent qemu (required for PowerPC to work)
- Add tests to run test.py for various QEMU platforms. This relies on
swarren's uboot-test-hooks repository to provide the abstractions.
Acked-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
Signed-off-by: Tom Rini <trini@konsulko.com>
We don't need to use TEST_CMD in order to run tests. We need a BUILDMAN
and TOOLCHAIN variable to avoid having to duplicate logic or write some
wrapper function. But this makes the tests harder as we add more
complex examples.
Signed-off-by: Tom Rini <trini@konsulko.com>
Acked-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
We can now build for microblaze, sh4 and xtensa.
Signed-off-by: Tom Rini <trini@konsulko.com>
Acked-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
- Drop the 'cache' line, travis-ci says to not cache apt packages (and
does not).
- Get the Ubuntu provided toolchain for ARM and PowerPC.
- Add more toolchain options that buildman can fetch.
Signed-off-by: Tom Rini <trini@konsulko.com>
Acked-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
We currently will always see a number of warnings due to device tree
issues. These (and other warnings) should not make the build be marked
as failure so catch exit status 129 specifically and return 0 in that
case.
Signed-off-by: Tom Rini <trini@konsulko.com>
Acked-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
Currently we fail to fetch the dtc.git tree due to an SSL issue within
the travis-ci environment. The easiest fix here is to switch to a git
URI.
Signed-off-by: Tom Rini <trini@konsulko.com>
Acked-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
In order to make other various improvements, update to the latest
environment travis-ci supports.
Signed-off-by: Tom Rini <trini@konsulko.com>
Acked-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
This provides runtime test coverage in Travis, in addition to the existing
build coverage.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Roger Meier <r.meier@siemens.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
Acked-by: Simon Glass <sjg@chromium.org>
Without this, builds default to using new Travis CI infra-structure which
does no allow sudo. The builds need sudo in order to install the ELDK
compilers. Consequently, almost all builds fail without this.
I suspect that existing Travis CI users have not noticed this because
their accounts or builds have been grand-fathered into backwards-
compatible default settings, whereas I just set up a new build from
scratch and received new default settings.
Cc: Wolfgang Denk <wd@denx.de>
Cc: Heiko Schocher <hs@denx.de>
Cc: Tom Rini <trini@ti.com>
Cc: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
Cc: Andreas Färber <afaerber@suse.de>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Acked-by: Heiko Schocher <hs@denx.de>
- add more targets for building with buildman:
- avr32
- m68k
and while at it, sort the list alphabetical
Reviewed-by: Roger Meier <r.meier@siemens.com>
Signed-off-by: Heiko Schocher <hs@denx.de>
