At present buildman does not write any output (to the 'out' and 'err)
files if the build terminates with a fatal error. This is to avoid adding
lots of spam to the logs.
However there are times when this is actually useful, such as when the
build fails for an obscure reason such as a Kconfig loop.
Update the logic to always write the output, so that the user gets a clue
as to what is happening.
Signed-off-by: Simon Glass <sjg@chromium.org>
At present we sometimes see problems in gitlab where the environment has
0x80 characters or sequences which are not valid UTF-8.
Avoid this by using bytes for the environment, both internal to buildman
and when writing out the 'env' file. Add a test to make sure this works
as expected.
Reported-by: Marek Vasut <marex@denx.de>
Fixes: e5fc79ea71 ("buildman: Write the environment out to an 'env' file")
Signed-off-by: Simon Glass <sjg@chromium.org>
There have been at least a few cases where an exception has occurred in a
thread and resulted in buildman hanging: running out of disk space and
getting a unicode error.
Handle these by collecting a list of exceptions, printing them out and
reporting failure if any are found. Add a test for this.
Signed-off-by: Simon Glass <sjg@chromium.org>
The environment may contain some unicode characters. At least that is what
seemed to happen on one commit:
Building current source for 1 boards (0 threads, 64 jobs per thread)
0 0 0 /1 -1 (starting)
Traceback (most recent call last):
File ".../tools/buildman/buildman", line 64, in <module>
ret_code = control.DoBuildman(options, args)
File "tools/buildman/control.py", line 372, in DoBuildman
options.keep_outputs, options.verbose)
File ".../tools/buildman/builder.py", line 1704, in BuildBoards
results = self._single_builder.RunJob(job)
File ".../tools/buildman/builderthread.py", line 526, in RunJob
self._WriteResult(result, job.keep_outputs, job.work_in_output)
File ".../tools//buildman/builderthread.py", line 349, in _WriteResult
print('%s="%s"' % (var, env[var]), file=fd)
UnicodeEncodeError: 'ascii' codec can't encode characters in position
311-312: ordinal not in range(128)
The problem defies repetition with any change at all to buildman. But
let's set an encoding in any case.
Signed-off-by: Simon Glass <sjg@chromium.org>
At present even if only a single thread is in use, buildman still uses
threading.
For some debugging it is helpful to do everything in the main process.
Allow -T0 to support this.
Signed-off-by: Simon Glass <sjg@chromium.org>
Buildman reuses build directories from previous builds to avoid the cost
of 'make mrproper' for every build. If the previous build produced an SPL
image but the current one does not, the SPL image will remain and buildman
will think it is a result of building the current board.
Remove these files before building, to avoid this problem.
Signed-off-by: Simon Glass <sjg@chromium.org>
At present patman sets the python path on startup so that it can access
the libraries it needs. If we convert to use absolute imports this is not
necessary.
Move patman to use absolute imports. This requires changes in tools which
use the patman libraries (which is most of them).
Signed-off-by: Simon Glass <sjg@chromium.org>
At present buildman does not write its own output files (err, done, the
environment) when using -w. However this is useful for when the build is
run with -s to check it.
In fact ProduceResultSummary() reads the result from those files rather
than using the 'result' info directly. So ProcessResult() does not work
with -w at present. It does not print any output.
Fix this by writing output files even when -w is used.
Signed-off-by: Simon Glass <sjg@chromium.org>
At present the environment used by U-Boot is written to the 'env'
directory. This is fine when the output directory is not the same as the
source directory, but when it is (as with -w) it conflicts with the source
directory of the same name.
Rename 'env' to 'out-env' to fix this.
Signed-off-by: Simon Glass <sjg@chromium.org>
At present buildman defaults to running 'mrproper' on every thread before
it starts building commits for each board. This can add a delay of about 5
seconds to the start of the process, since the tools and other invariants
must be rebuilt.
In particular, a build without '-b', to build current source, runs much
slower without -I, since any existing build is removed, thus losing the
possibility of an incremental build.
Partly this behaviour was to avoid strange build-system problems caused by
running 'make defconfig' for one board and then one with a different
architecture. But these problems were fixed quite a while ago.
The -I option (which disabled mrproper) was introduced four years ago and
does not seem to cause any problems with builds.
So make -I the default and deprecate the option. To allow use of
'mrproper', add a new -m flag.
Signed-off-by: Simon Glass <sjg@chromium.org>
It is useful to run a simple build and put all the output in a single
directory. Add a -w option to support this.
Signed-off-by: Simon Glass <sjg@chromium.org>
This class has a few more members now. Add documentation for them and fix
a nit in the 'commits' comment.
Signed-off-by: Simon Glass <sjg@chromium.org>
Sometimes it is useful to see the environment that was used to build
U-Boot. Write this out to a file in the build directory.
Signed-off-by: Simon Glass <sjg@chromium.org>
The 'done' files created by buildman may end up being empty if buildman
runs out of disk space while writing them. This error is then persistent,
since even if disk space is reclaimed and the build retries, the empty
file causes an exception in the builder thread.
Deal with this silently by doing a rebuild.
Signed-off-by: Simon Glass <sjg@chromium.org>
As we're building the boards, extract the default U-Boot environment to
uboot.env so we can interrogate it later.
Signed-off-by: Alex Kiernan <alex.kiernan@gmail.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
When the U-Boot base directory happens to have the same name as the branch
that buildman is directed to use via the '-b' option and no output
directory is specified with '-o', buildman happily starts removing the
whole U-Boot sources eventually only stopped with the error message:
OSError: [Errno 20] Not a directory: '../<branch-name>/boards.cfg
Add a check to avoid this and also deal with the case where '-o' points
to the source directory, or any subdirectory of it.
Finally, tidy up the confusing logic for removing the old tree when using
-b. This is only done when building a branch.
Signed-off-by: Lothar Waßmann <LW@KARO-electronics.de>
Signed-off-by: Simon Glass <sjg@chromium.org>
Tested-by: Lothar Waßmann <LW@KARO-electronics.de>
When U-Boot started using SPDX tags we were among the early adopters and
there weren't a lot of other examples to borrow from. So we picked the
area of the file that usually had a full license text and replaced it
with an appropriate SPDX-License-Identifier: entry. Since then, the
Linux Kernel has adopted SPDX tags and they place it as the very first
line in a file (except where shebangs are used, then it's second line)
and with slightly different comment styles than us.
In part due to community overlap, in part due to better tag visibility
and in part for other minor reasons, switch over to that style.
This commit changes all instances where we have a single declared
license in the tag as both the before and after are identical in tag
contents. There's also a few places where I found we did not have a tag
and have introduced one.
Signed-off-by: Tom Rini <trini@konsulko.com>
Add a new option '-E' for treating all compiler warnings as errors.
Eventually this will pass 'KCFLAGS=-Werror' to Kbuild.
Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
The build output can still produce unicode encoded output. But in
the buildman's log and err files we only want plain ASCII characters.
To handle all situations with unicode and non-unicode output, encode
the stdout and stderr strings to UTF-8 and afterwards to ASCII with
replacing all special characters.
Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
When writing out some of our results we may now have UTF-8 characters
in there as well. Translate these to latin-1 and ignore any errors (as
this is for diagnostic and given the githash anything else can be
reconstructed by the user.
Reviewed-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Tom Rini <trini@konsulko.com>
Normally buildman does a full build of a board. This includes creating the
u-boot.cfg file which contains all the configuration options. Buildman uses
this file with the -K option, to show differences in effective configuration
for each commit.
Doing a full build of U-Boot just to create the u-boot.cfg file is wasteful.
Add a -D option which causes buildman to only create the configuration. This
is enough to support use of -K and can be done much more quickly (typically
5-10 times faster).
Signed-off-by: Simon Glass <sjg@chromium.org>
Signed-off-by: York Sun <york.sun@nxp.com>
CC: Simon Glass <sjg@chromium.org>
Fixed commit subject:
Signed-off-by: Simon Glass <sjg@chromium.org>
Acked-by: Simon Glass <sjg@chromium.org>
One use-case for buildman is to continually run it interactively after
each small step in a large refactoring operation. This gives more
immediate feedback than making a number of commits and then going back and
testing them. For this to work well, buildman needs to be extremely fast.
At present, a couple issues prevent it being as fast as it could be:
1) Each time buildman runs "make %_defconfig", it runs "make mrproper"
first. This throws away all previous build results, requiring a
from-scratch build. Optionally avoiding this would speed up the build, at
the cost of potentially causing or missing some build issues.
2) A build tree is created per thread rather than per board. When a thread
switches between building different boards, this often causes many files
to be rebuilt due to changing config options. Using a separate build tree
for each board would avoid this. This does put more strain on the system's
disk cache, but it is worth it on my system at least.
This commit adds two command-line options to implement the changes
described above; -I ("--incremental") turns of "make mrproper" and -P
("--per-board-out-dir") creats a build directory per board rather than per
thread.
Tested:
./tools/buildman/buildman.py tegra
./tools/buildman/buildman.py -I -P tegra
./tools/buildman/buildman.py -b tegra_dev tegra
./tools/buildman/buildman.py -b tegra_dev -I -P tegra
... each once after deleting the buildman result/work directory, and once
"incrementally" after a previous identical invocation.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Tom Rini <trini@konsulko.com>
Acked-by: Simon Glass <sjg@chromium.org> # v1
Tested-by: Simon Glass <sjg@chromium.org> # v1
Acked-by: Simon Glass <sjg@chromium.org>
On i.MX platforms the SPL binary is called "SPL" so make sure we keep
that.
Cc: Simon Glass <sjg@chromium.org>
Signed-off-by: Tom Rini <trini@konsulko.com>
Acked-by: Simon Glass <sjg@chromium.org>
The help text for -V says we will pass V=1 but all it really did was not
pass in -s. Change the logic to pass make V=1 with given to buildman -V or
-s to make otherwise.
Cc: Simon Glass <sjg@chromium.org>
Signed-off-by: Tom Rini <trini@konsulko.com>
Acked-by: Simon Glass <sjg@chromium.org>
When told to keep outputs, be much more liberal in what files we keep.
In addition to adding 'MLO', keep anything that matches u-boot-spl.* (so
that we keep the map file as well) and anything we generate about
'u-boot itself. A large number of bootable formats now match this and
thus it's easier to build many targets and then boot them afterwards
using buildman.
Cc: Simon Glass <sjg@chromium.org>
Signed-off-by: Tom Rini <trini@konsulko.com>
Acked-by: Simon Glass <sjg@chromium.org>
At present buildman tries to detect an aborted build and doesn't record a
result in that case. This is to make sure that an abort (e.g. with Ctrl-C)
does not mark the build as done. Without this option, buildman would never
retry the build unless -f/-F are provided. The effect is that aborting the
build creates 'fake errors' on whatever builds buildman happens to be
working on at the time.
Unfortunately the current test is not reliable and this detection can
trigger if a required toolchain tool is missing. In this case the toolchain
problem is never reported.
Adjust the logic to continue processing the build result, mark the build as
done (and failed), but with a return code which indicates that it should be
retried.
The correct fix is to fully and correctly detect an aborted build, quit
buildman immediately and not write any partial build results in this case.
Unfortunately this is currently beyond my powers and is left as an exercise
for the reader (and patches are welcome).
Signed-off-by: Simon Glass <sjg@chromium.org>
Normally buildman runs with 'make -s' meaning that only errors and warnings
appear in the log file. Add a -V option to run make in verbose mode, and
with V=1, causing a full build log to be created.
Signed-off-by: Simon Glass <sjg@chromium.org>
In some cases there may be multiple toolchains with the same name in the
path. Provide an option to use the full path in the CROSS_COMPILE
environment variable.
Note: Wolfgang mentioned that this is dangerous since in some cases there
may be other tools on the path that are needed. So this is set up as an
option, not the default. I will need test confirmation (i.e. that this
commit fixes a real problem) before merging it.
Signed-off-by: Simon Glass <sjg@chromium.org>
Suggested-by: Steve Rae <srae@broadcom.com>
When saving binary files we likely want to keep any .img files that have
been generated as well.
Signed-off-by: Tom Rini <trini@ti.com>
Acked-by: Simon Glass <sjg@chromium.org>
When creating build directories also create parents as necessary. This
fixes a failure when building a hierarchical branch (i.e. foo/bar).
Signed-off-by: Thierry Reding <treding@nvidia.com>
Acked-by: Simon Glass <sjg@chromium.org>
Tested-by: Tom Rini <trini@ti.com>
When a build is to be performed, buildman checks to see if it has already
been done. In most cases it will not bother trying again. However, it was
not reading the return code from the 'done' file, so if the result was a
failure, it would not be counted. This depresses the 'failure' count stats
that buildman prints in this case.
Fix this bug by always reading the return code.
Signed-off-by: Simon Glass <sjg@chromium.org>
The full path is long and also includes buildman private directories.
Clean this up, so that only a relative U-Boot path is shown.
This will change warnings like these:
/home/sjg/c/src/third_party/u-boot/buildman5/.bm-work/00/arch/sandbox/cpu/cpu.c: In function 'timer_get_us':
/home/sjg/c/src/third_party/u-boot/buildman5/.bm-work/00/arch/sandbox/cpu/cpu.c:40:9: warning: unused variable 'i' [-Wunused-variable]
/home/sjg/c/src/third_party/u-boot/files/arch/sandbox/cpu/cpu.c: In function 'timer_get_us':
/home/sjg/c/src/third_party/u-boot/files/arch/sandbox/cpu/cpu.c:40:9: warning: unused variable 'i' [-Wunused-variable]
to:
arch/sandbox/cpu/cpu.c: In function 'timer_get_us':
arch/sandbox/cpu/cpu.c:40:9: warning: unused variable 'i' [-Wunused-variable]
Signed-off-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Roger Meier <roger@bufferoverflow.ch>
Reviewed-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Tested-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Cc: Tom Rini <trini@ti.com>
Cc: Wolfgang Denk <wd@denx.de>
Cc: Simon Glass <sjg@chromium.org>
