| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
| |
Some tests were removed in commit f4d3cbaf1d6e5c3bdbb1b2e2 but it seems
there are still code remnants that we don't need any more. There may be
more -- I haven't done a thorough job of this.
Change-Id: I0d7151402365f9e6f23a93a1d30df0198de9ef02
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When preparing the build directory for a chunk, Morph clones the Git repo
being built (and any submodules) into the staging area.
Instead of using `git clone --no-hardlinks`, the
morphlib.git.copy_repository() function uses 'cp -a' plus some faffing
to achieve the same thing a bit faster.
The 'cp -a' command will preserve the ownership and permissions of the
cached copy. These may not match up with which user is doing the build.
I found that files in my Git cache were owned by UID 1002. This caused
chunks that needed to access the .git directory at build time to fail
with strange errors. Worse, it would trigger a bug in Git[1] that leads
to a fork bomb, which would cause Linux to freeze up completely. This
occured even though I was building as `root`, because of the way
`linux-user-chroot` drops certain privileges: presumably, setting
SECBIT_NOROOT leads to CAP_DAC_OVERRIDE being unset.
To avoid this bug, the code now ensures the copied .git repos are owned
by the user and group who ran `morph`.
Another way to fix this would be to change the
morphlib.gits.copy_repository() function to use `git clone
--no-hardlinks`. This is what YBD does. I found that there is a slight
speed benefit to using the current code ... these are results of cloning
repos from the cache with the 2 methods:
With 'cp -a' plus a chown:
upstream:bison: 4.54 seconds average (10 results)
upstream:ybd: 0.13 seconds average (10 results)
upstream:linux: 40.51 seconds average (10 results)
With 'git clone --hardlinks':
upstream:bison: 6.23 seconds average (10 results)
upstream:ybd: 0.11 seconds average (10 results)
upstream:linux: 43.36 seconds average (10 results)
Test code is: <https://gist.github.com/ssssam/833e0ef8d04fb1fb6ff3>.
Ideally we would fix `git clone --no-hardlinks` to be faster, but we
may as well keep the existing code for the time being.
[1]: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=750687
Change-Id: Ieea87322ea7b7f62975b9480f877755665656217
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Loading .morph files is becoming a bit more complicated, as we need to
deal with the VERSION file, and possibly soon with a DEFAULTS file as
well.
The logic of loading and parsing .morph files is done either in the
sourceresolver module, or the morphloader module. This change means that
all users of the latter module can use the get hold of a
MorphologyLoader instance with VERSION already parsed. If DEFAULTS is
added then it is also simple to parse DEFAULTS.
Change-Id: Ib33756e9dbd078e38f12dd7f776c89584b178959
|
|
|
|
|
|
|
| |
Using app.status() produces nicer logs as the messages will have the
status prefix prepended.
Change-Id: I0573d7066784afdbfb878bed85af704e82bb9797
|
|
|
|
|
|
|
|
|
| |
Replace cliapp.shell_quote() with pipes.quote() to produce the chroot
scripts. Despite pipes.quote() being deprecated, it produces much less
escaping than cliapp.shell_quote(). This method should be replaced with
shlex.quote() when migrating to Python 3.
Change-Id: I4b8e8eefe0cd321458bb3fae72f6d0552680f84f
|
|
|
|
| |
Change-Id: I276a00bb9117aaa2d4fcc7e187b846eedae65591
|
|
|
|
|
|
| |
... instead the more common autogen[.sh]
Change-Id: I3d797581708ffaefd1a1d4880253cfc2a3f5aa85
|
|
|
|
|
|
|
|
|
|
| |
Commit a4d8098b229592db40565747b0444b518bf8a6eb removed the `morph
foreach` command, which broke some `morph edit` tests. I don't think
we are ready to remove `morph edit` yet (documentation still mentions
it) but we can remove these tests, which require `morph foreach` to
work.
Change-Id: Id372760f69f31a46d1d872ee36ec6fdfb324ce34
|
|
|
|
|
|
|
|
|
| |
Looks like,
2015-07-05 16:08:10 [Build 1/304] [stage1-binutils] Fetching to local cache: artifact stage1-binutils-misc
stage1-binutils-misc[##################### ] 51.9/73.0 MB
Change-Id: Ib10f1cfaa0c1df80ae605ecfeb5b706c8d46c4a4
|
|
|
|
| |
Change-Id: Iedefd1745e85e1f30d45ee8d26b315760a542a0b
|
|
|
|
|
|
|
|
|
|
|
|
| |
Pass the Source to the staging area constructor so that we don't need to
pass it as a parameter when we call some StagingArea methods. Also move
the creation of the build and destdir directories to the constructor so
that we can get rid of the chroot_open() and chroot_close() methods.
Also, provide API to retrieve the relative locations for buildir and
destdir.
Change-Id: I6e8085392e19ff3d8df807f260acf90eec9e0901
|
|
|
|
| |
Change-Id: Iad95af65bd5c528d2e72f5b2ffa80a01152f50ff
|
|
|
|
|
|
|
|
| |
We're beginning a transition away from workspaces,
foreach in particular is possibly one of the least used commands,
and is only mentioned in tutorials explicitly marked as 'old'.
Change-Id: I2c95ce689bb5550bf50feb6b55be6c58671c4c4a
|
|
|
|
| |
Change-Id: Icec7e948c60decefe5a8bab6141bc89603c265b7
|
|
|
|
| |
Change-Id: I6c45763fb4d9f304695c1894625ced36bda36f89
|
|
|
|
|
|
|
|
|
| |
This change, suggested in story #16, causes the testing suite to use
the user-defined tempdir in /etc/morph.conf if it exists or continue
using /tmp if not. This will prevent the test suite from filling drives
that are not intended to be used for development work.
Change-Id: I94edabe4ea1eca7a55ece7424ece7bbd4b68c513
|
|
|
|
| |
Change-Id: I9e5b1a70946a87eb29009df2def3bd98ecc7ad2a
|
|
|
|
|
|
|
|
| |
ExtUtils::MakeMaker is preferred, Module::Build was meant to replace it
but essentially wasn't good enough, some projects still use Module::Build
though.
Change-Id: I124ee7b33f32167302e9bcb5299f6422f4fc346e
|
|
|
|
|
|
| |
Install to default locations rather than overriding.
Change-Id: Ieccd06afdf623eee0d09b8d29d40d147dc8c5e7c
|
|
|
|
|
|
| |
There are several perl build systems.
Change-Id: I17b5d61b7bffbdbcb9944e9e41449bec7a69c527
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Improve error messages and remove redundant information
* Stop displaying the containerised settings, as they are not very
helpful to debug a build failure
* Display the error code
* Change system integration commands logging to consistent with build
commands logging.
* Provide a log file in the staging directory for each chunk build
failure or system integration failure.
* Provide a chroot script in the staging directory for each chunk build
failure or system integration build failure. This script allows the
user to enter an environment similiar to the one where the build
failed.
* Fix a bug where the system integration commands stdout was being
omitted when these commands failed.
* Fix a bug where the error_message_for_containerised_commandline() was
being called with an empty error message, as the stderr was being
directed to stdout.
Chunk build failure output before the this change and change
I16ebe9ba4ac46fef82e37d0b3e05f42d14249de8:
2015-08-03 15:55:21 [Build 258/267] [test] Running build-commands
build failed
# configure
# # echo "configure-commands stdout"
configure-commands stdout
# # echo "configure-commands stderr" >&2
configure-commands stderr
# build
# # echo "build commands stdout"
build commands stdout
# # echo "build commands stderr" >&2
build commands stderr
# # echo "some more sdout"
echo "some more stderr" >&2
echo $PATH
foo
some more sdout
some more stderr
/tools/bin:/usr/bin:/usr/lib/ccache:/sbin:/usr/sbin:/bin:/usr/bin
sh: foo: not found
ERROR: In staging area /src/tmp/failed/tmpglc5Sz: Command failed: sh -c
'echo "some more sdout"
echo "some more stderr" >&2
echo $PATH
foo
':
Containerisation settings: {'binds': (('/src/cache/ccache/nano-tarball',
'/src/tmp/staging/tmpglc5Sz/tmp/ccache'),), 'mount_proc': True,
'mounts': (('dev/shm', 'tmpfs', 'none'),), 'writable_paths':
['/src/tmp/staging/tmpglc5Sz/test.build',
'/src/tmp/staging/tmpglc5Sz/test.inst',
'/src/tmp/staging/tmpglc5Sz/dev', '/src/tmp/staging/tmpglc5Sz/proc',
'/src/tmp/staging/tmpglc5Sz/tmp'], 'root': '/src/tmp/staging/tmpglc5Sz',
'cwd': '/test.build'}
Error output:
$>
Chunk build failure output after this change and change
I16ebe9ba4ac46fef82e37d0b3e05f42d14249de8:
2015-08-03 15:53:59 [Build 258/267] [test] Running build-commands
### CONFIGURE-COMMANDS ###
+ echo configure-commands stdout
configure-commands stdout
+ echo configure-commands stderr
configure-commands stderr
### BUILD-COMMANDS ###
+ echo build commands stdout
build commands stdout
+ echo build commands stderr
build commands stderr
+ echo some more sdout
some more sdout
+ echo some more stderr
some more stderr
+ echo /tools/bin:/usr/bin:/usr/lib/ccache:/sbin:/usr/sbin:/bin:/usr/bin
/tools/bin:/usr/bin:/usr/lib/ccache:/sbin:/usr/sbin:/bin:/usr/bin
+ foo
sh: foo: not found
ERROR: In staging area /src/tmp/staging/tmpjES5_2: build failed (exi
t_code=127)
$>
System integration build failure output before this change and change
I16ebe9ba4ac46fef82e37d0b3e05f42d14249de8:
2015-08-03 15:58:25 [Build 267/267] [build-system-x86_64] Running the
system integration commands
2015-08-03 15:58:25 [Build 267/267] [build-system-x86_64] Error while
running system integration commands
2015-08-03 15:58:25 [Build 267/267] [build-system-x86_64] Error while
building system
ERROR: Command failed:
baserock/system-integration/00-test-test-misc-0002:
Containerisation settings: {'mounts': (('dev/shm', 'tmpfs', 'none'),
('tmp', 'tmpfs', 'none')), 'mount_proc': True, 'root':
'/src/tmp/staging/tmpvpwB4l/build-system-x86_64.inst'}
Error output:
+ foo
baserock/system-integration/00-test-test-misc-0002: line 1: foo: not
found
$>
System integration build failure output after this change and change
I16ebe9ba4ac46fef82e37d0b3e05f42d14249de8:
2015-08-03 16:00:06 [Build 267/267] [build-system-x86_64] Running the
system integration commands
+ echo stdout
stdout
+ echo stderr
stderr
+ foo
baserock/system-integration/00-test-test-misc-0002: line 1: foo: not
found
ERROR: In staging area /src/tmp/staging/tmpl9VNzf: system integration
commands failed (exit_code=127)
$>
Change-Id: Id992f707f69f3fa761b4c21e9904c4e5328e1c77
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Stop moving staging areas of failed builds from the 'staging' directory
to the 'failed' directory. Moving staging areas make it very difficult
to debug build failures on the build essential chunks, as the paths set
on the configure scripts and some environment variables (e.g.
STAGE2_SYSROOT, DESTDIR) will be invalid after moving the staging area.
This change will also make it easier to create scripts that chroot n
environment similiar to the one where the build failure occurred.
To make it still possible to safely do a build an run `morph gc` in
parallel, we use flock(2) to control access to the staging area
directory.
Also, move the `test_supports_non_isolated_mode` test into a different
class, as it requires a different SetUp() routine (the staging area is
contructed with different parameters).
Change-Id: I06c3c435ad05c12afabc0adc2a9d4f8a284ccc02
|
|
|
|
| |
Change-Id: I2e1ae43da6e7dd0d7ddcb51d3cd9bc1794bb80a5
|
|
|
|
|
|
|
| |
I need to use this outside of the 'sourceresolver' module. Also, the
'sourceresolver' module is too big.
Change-Id: I523bcc9555193b7369768441b72f1059e6adde5c
|
|
|
|
|
|
| |
Just for future debuggers.
Change-Id: Ie915a9f47fbf3ca56d6e75b816eaa5a4fb36bc88
|
|
|
|
|
|
|
|
|
|
|
| |
Instead of logging every command for ourselves, use the shell X-ray
mode. As side effect, each command argument will now be printed out in
its evaluated form, which aids debugging.
Also, improve the visual separation between the different build steps,
and display those steps with finer granularity.
Change-Id: I16ebe9ba4ac46fef82e37d0b3e05f42d14249de8
|
|
|
|
|
|
|
| |
This is probably a sign that something is broken, but it's better to
cope with it rather than crashing.
Change-Id: I4e9c40d0b7db031ea271143938db61c4ad70a888
|
|
|
|
|
|
|
|
|
| |
When I spotted that artifacts contained a .git directory I was worried
that it might indicate a Morph bug. It turned out to be due to the way
the 'test shell' used in the Yarn tests works. This change should save
anyone else from similar confusion.
Change-Id: I97b57eff7aa947506d4062a76e9803398e46c2b6
|
|
|
|
|
|
|
|
|
|
|
|
| |
The system-integration field in a chunk .morph file defines which of the
chunk artifacts the integration script should live. If you list an
artifact that doesn't exist, the integration script will not go anywhere
and the commands won't run.
Ideally system-integration commands will be become a bit more simple to
use, but hopefully this warning helps in the meantime.
Change-Id: I10a068e79eb46f98f0bd308a5caf3c445dda22ed
|
|
|
|
|
|
|
| |
Chunk commands will currently block if they read from stdin.
To fix this we set stdin to /dev/null so that the first read returns EOF.
Change-Id: I143396f4c443c098b2880c5925e6f0907cb4fb17
|
|
|
|
|
|
| |
They were all run onto the same line before with no linebreaks.
Change-Id: Ibdff91a23221034a3d345542f1268f8863cf4515
|
|
|
|
| |
Change-Id: I891d1b13ed0581b293fe6b09b3cc73af8fd81d67
|
|
|
|
|
|
|
|
|
|
|
| |
The previous release of the Baserock reference systems (baserock-15.25
tag) used Baserock definitions version 3, so this version of Morph can
still build the last release of the Baserock reference systems.
In the course of fixing up the cmdtest tests to use version 5
definitions, I removed a couple of bits of redundant code.
Change-Id: If9d480aa966d2ea5939cf04351f3d488cccea67f
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This makes more sense to me conceptually: the idea of the 'manual' build
system is you specify *everything* manually.
However, I made this change to work around a problem in the Yarn test
suite. If the definitions used for testing are set to version 5 or
newer, `morph build` will try to run 'strip' and 'objcopy', which fails
because they don't exist in the test definitions.
We could solve this by setting 'strip-commands: []' in the test chunk
.morph files, but this triggers an error further down where `morph edit`
fails with 'Field strip-commands not allowed in morphology
test-chunk.morph'. This bug is caused by the conditional way that
strip-commands are currently implemented in Morph. When we drop support
for versions of the definitions format older than V5, the bug will go
away, and the `morph edit` command is deprecated in any case, so I think
it makes sense to work around it instead of trying to fix it.
Change-Id: Ib5124d72466a77cf3f28ed3e14cc4c231bdce4c4
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
tar will preserve ownership and permissions by default
if running as the superuser, whereas git doesn't track ownership,
the owner is set to the user running the clone.
This is a reproducibility hole since the resulting repo may differ
depending on whether the repo was obtained from a tarball or via a git
clone, but morph considers the repos to be equivalent.
Running tar with --no-same-owner ensures the files
written by tar will be owned by the user extracting the archive.
This only fixes half the problem,
lorry may also need to be modified to ensure files in the
tarball have the same mode as those in the git repo.
Change-Id: I849d054bc4574f9c0bbcfd608914306e52b7a22e
|
|
|
|
| |
Change-Id: Ife16546fc90e26919672f6658b30084bafa42f94
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This test, tests for a case that produces unhelpful results.
It asserts that if we don't have the repository locally, and we try to
fetch it remotely and the file doesn't exist, that we should have no
morphology.
What we'd actually like is for it to fetch the repository and try that
locally, at which point it would succeed.
The reason the flawed previous behaviour (that this test asserted to be
correct) failed to cause problems for us, is that in the normal course
of resolving refs, we would always either have found it remotely or
cached it locally, when we were doing ls_tree or resolve_ref operations.
Change-Id: I287eb1e5bc9c916b31f306b4f5f24b0a6d3f0559
|
|
|
|
|
|
|
| |
If we did end up trying to load a morphology that wasn't defined, we'd
get an unhelpful error about the yaml being invalid.
Change-Id: If1b182f1100430424a0fe62bd8938a88c1b9a11a
|
|
|
|
|
|
|
|
| |
The initiator would always say "Need to build 260/260 artifacts" even
when it didn't need to build everything, because we were counting the
number of unbuilt artifacts wrongly.
Change-Id: I5da88157dba59949597c58a983f7b31975c52d7f
|
|
|
|
|
|
|
| |
Bad function prototype meant that the mechanism for handling workers
disconnecting actually caused the controller to crash instead.
Change-Id: I8ceb6ad027ba2481c0c4c335e1760692823c208b
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
I was getting an error from this command:
morph distbuild systems/build-system-x86_64.morph stage2-fhs-dirs
Saying this:
ERROR: Failed to build
git://git.baserock.org/baserock/baserock/definitions
93575a2ceeeda77a5bb8c6121a9cac3edde1afbf
systems/build-system-x86_64.morph: Some of the requested components
are
not in build-system-x86_64-rootfs: stage2-fhs-dirs
This patch fixes that issue and makes Morph build up to stage2-fhs-dirs
successfully.
Change-Id: I61c373272484dcb5dc62f281cae8f21f742c31a9
|
|
|
|
|
|
|
|
| |
This is just to make the log files more readable, as what would
previously have been logged as '<ArtifactReference at 0x1235478>' is
now logged as the actual name of the artifact.
Change-Id: I6189aa1390268cec379dd459fc3f4fecc71363b1
|
|
|
|
|
|
|
|
| |
The 'BC: got artifact: <distbuild.artifact_reference.ArtifactReference
object at 0x7f84ea2b5c10>' message is only useful when
_debug_build_output is true, if at all.
Change-Id: I079b398e841d5508ecefd00167fb0d83be748ce6
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This fixes an issue where distbuild would build the same artifact more
than once. The problem occurs with a single distbuild controller, if
multiple initiators request builds of the same thing at roughly the
same time (which scripts/release-build in definitions.git does).
This change also means that multiple distbuild controllers sharing a
single artifact cache will be smart about sharing built artifacts. It
does not mean that distbuild can handle having built artifacts removed
from the cache while it is building stuff.
The number of HTTP requests made to the shared artifact cache is higher
with this patch, but these seem to take no more than 1 second and we
only ever need to run one request before starting more builds, so there
should be no noticable impact on performance.
Change-Id: Ib3246219a10ca95d40b8a21bd0fe53f32e46c1c9
|
|
|
|
|
|
|
| |
Hopefully this makes the code a little less cryptic. No functional
changes.
Change-Id: I615810e4eacdd5454731e07387b1dbb9eb348fd5
|
|
|
|
|
|
|
|
| |
This was causing:
UnboundLocalError: local variable 'original_ref' referenced before assignment
Change-Id: I46e4b5d527d9ac5480a3d1fe5e6d631f0b7279dc
|
|
|
|
|
|
|
| |
This makes it more useful and saves people from poking around in the
cache in order to dig up build logs.
Change-Id: I8e062c5c32b01aca0df54e1974ead3c3b3134cc3
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This makes `morph get-chunk-details`, `morph get-repo` and `morph
show-build-log` work in any Git checkout of definitions.git. These
are all of the commands outside of the deprecated
branch_and_merge_plugin that take notice of system branches.
The DefinitionsRepo.relative_path_to_chunk() function is changed a
bit. It actually only existed for the `get-repo` command to use, but
it turns out that it did the wrong thing by returning a long path for
the repo (e.g. baserock/baserock/fhs-dirs) instead of a short one
(fhs-dirs). The latter is less typing, and is the behaviour expected
by the `get-repo` yarn tests, so it now does that.
Change-Id: I430b540b3b0f309cf7018e0b8236f0e8a9042d89
|
|
|
|
|
|
|
| |
Previously callers to DefinitionsRepoWithApp.source_pool() could
override the user's value, but that's not actually used anywhere.
Change-Id: I43b53e7b9fc937886c8e6e95947e5e2b6776d085
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Previously it would print nothing and then exit with success. Better
to tell the user that an error occurred.
Example of the new error:
ERROR: No build log for artifact
6fb39673b8f9a1c9848063f5132aa958ffa75f2be61a8dde68ebb11f5a3c4a5f found
on cache server http://cache.baserock.org:8080/
Change-Id: Icf8ceef60eb497ff90e00391d442a394e3d76f10
|