| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
| |
Updates the refs in the job process (but doesn't write), to ensure
following sources can see consistency of previous sourcse has been
updated. `_save_ref` is renamed `_set_ref` with writing to file now
optional.
This also changes the previous_source_access test to use a remote, so
that it actually tests this cornercase.
Fixes #1010
|
|\
| |
| |
| |
| | |
jobs/job.py: refactor, rm redundant _retry_flag
See merge request BuildStream/buildstream!1335
|
|/
|
|
|
| |
This member variable is always assigned right before it's tested, so it
doesn't need to persist in the object.
|
|\
| |
| |
| |
| |
| |
| | |
Fix build track all no strict
Closes #1014
See merge request BuildStream/buildstream!1332
|
| |
| |
| |
| |
| |
| | |
actually build
This is a regression test for issue #1014
|
| |
| |
| |
| |
| | |
This adds a check to test_build_track() to ensure that the target
is cached as a result of building with tracking of selected elements.
|
|/
|
|
|
|
|
|
|
|
|
|
|
| |
When calculating the strict key, we do not know for sure yet that the
element is required or whether it will ever have a cache key, as
we could be tracking elements which will result in resolving an
element which is a build-only dependency of the target.
This ensures that we process all elements which need to be processed,
which is important for cases where we are building in non-strict mode
and tracking is also enabled.
This fixes issue #1014
|
|\
| |
| |
| |
| |
| |
| | |
Fix no strict junctions
Closes #1018
See merge request BuildStream/buildstream!1333
|
| |
| |
| |
| |
| |
| | |
non-strict mode
This is a regression test for #1018
|
|/
|
|
|
|
|
|
|
|
|
|
| |
Recent code changes has avoided resolving cache keys in some conditions,
which causes loading to break when attempting to stage junctions in non-strict
mode.
Instead of calling Element._update_state() here to resolve the
cache key, setting it to be required causes it to automatically
have it's cache key resolved, solving the problem.
This change fixes #1018
|
|\
| |
| |
| |
| |
| |
| | |
element.py: Reset workspace state if last successful build is missing.
Closes #1017
See merge request BuildStream/buildstream!1328
|
| |
| |
| |
| |
| |
| | |
artifacts are deleted
This is a regression test for #1017
|
| |
| |
| |
| |
| |
| |
| | |
The artifact directories are based on the element normal_name, which
substitutes any path separators with dashes.
Fix the helper function to use the correct path.
|
|/
|
|
|
|
|
|
| |
If the artifact referred to by the last successful build in the
workspace state no longer exists, reset the workspace state and
do not attempt to perform an incremental build instead of crashing.
This fixes #1017
|
|\
| |
| |
| |
| | |
tox.ini: do not hardcode sphinx version
See merge request BuildStream/buildstream!1306
|
|/
|
|
|
| |
This is not needed anymore since
431e578bf1b80074bd0f0ca8559e998dad59d4c0
|
|\
| |
| |
| |
| | |
.gitlab-ci.yml: Add Fedora 30 and remove deprecated Fedora 28
See merge request BuildStream/buildstream!1320
|
|/ |
|
|\
| |
| |
| |
| | |
_yaml.py, source.py: Fix tracking of refs inside conditions
See merge request BuildStream/buildstream!1324
|
|/
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Previously if you had a source of the form:
```yaml
- (?):
somecondition:
url: blah
ref: foo
othercondition:
url: blah
ref: bar
```
And you did `bst source track` on the element then you'd get something
output of the form:
```yaml
- (?)...
...
ref: wibble
```
With this patch, the *correct* ref inside the conditionals is updated
instead.
Signed-off-by: Daniel Silverstone <daniel.silverstone@codethink.co.uk>
|
|\
| |
| |
| |
| |
| |
| | |
_artifactcache.py: display the actual key being pulled
Closes #981
See merge request BuildStream/buildstream!1304
|
|/
|
|
| |
Fixes #981
|
|\
| |
| |
| |
| | |
_yaml.py: remove duplicated check
See merge request BuildStream/buildstream!1318
|
|/ |
|
|\
| |
| |
| |
| |
| |
| | |
Fix cloned plugin ids
Closes #1012
See merge request BuildStream/buildstream!1316
|
| |
| |
| |
| |
| |
| | |
When cloning a Source, we should inherit the same unique ID for
the sake of sending a valid ID in any messages sent back to
the frontend from a source cloned in a child task.
|
| |
| |
| |
| |
| |
| |
| | |
In the case of cloned Sources, they should not be allocating
a new ID in track() and fetch() in case they do communicate their
ID back to the main process, they should inherit the same ID of
the Source they were cloned from.
|
|/
|
|
|
|
|
|
|
|
|
| |
This was always intended, but was not well commented. The reason
we start plugin ID counters at 1 is that we prefer relying on
a falsy value to determine whether an ID holding variable has
been set or not.
This patch also adds a more informative assertion in Plugin._lookup()
This by itself essentially fixes #1012
|
|\
| |
| |
| |
| | |
testing._sourcetests: Don't special case 'local' in parameter list
See merge request BuildStream/buildstream!1317
|
|/
|
|
|
|
|
|
|
|
|
|
| |
The 'local' kind is hard coded in the parameter list of on test in
testing._sourcetests as so will always run regardless of what plugins
have been registered.
Remove this special casing by duplicating the test in the local source
specific test.
Ideally, the local source should have a Repo implementation and be
registered with the templated tests.
|
|\
| |
| |
| |
| | |
Reduce the amount of times we call Element._update_state()
See merge request BuildStream/buildstream!1312
|
| |
| |
| |
| | |
Nothing can happen at this point
|
|/ |
|
|\
| |
| |
| |
| | |
tests/testutil/python_repo.py: fixed executable path when running pip
See merge request BuildStream/buildstream!1313
|
|/
|
|
|
| |
Using `sys.executable` is a safer option as that is the interpreter
that is currently being used
|
|\
| |
| |
| |
| | |
Make Integration cache fixture available in testing package
See merge request BuildStream/buildstream!1307
|
|/
|
|
|
|
| |
The cli_integration fixture provided in testing.runcli depends on the
integration cache fixture. This was missed when cli_integration was
originally exposed.
|
|\
| |
| |
| |
| | |
Simplify metadata within Artifact class
See merge request BuildStream/buildstream!1314
|
| |
| |
| |
| |
| |
| | |
The returned key was only being used to reference or create the
keyed dicts within the metadata methods. Keeping key as a default
optional parameter may later be found redundant.
|
| |
| |
| |
| |
| |
| | |
As with the metadata simplification, the key is internalised to the
element's Artifact member as such it is redudant to return the key
which was used.
|
|/
|
|
|
|
| |
As the Artifact instance within element represents a single artifact
with the keys internalised, this removes the need to have keyed dicts
for the extracted artifact metadata.
|
|\
| |
| |
| |
| | |
.gitlab-ci.yml: Updating overnight tests to test ported freedesktop-sdk
See merge request BuildStream/buildstream!1308
|
|/
|
|
|
|
|
|
|
| |
Now that BuildStream 2 APIs have diverged, we maintain a branch of
freedesktop-sdk which supports BuildStream 2 and uses bst-plugins-experimental
instead of bst-external.
This commit updates our overnight tests to check that we're not breaking
our builds against a ported freedesktop-sdk project.
|
|\
| |
| |
| |
| | |
Revert !1164 - Manage GC during pipeline load
See merge request BuildStream/buildstream!1310
|
|/
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This change was initialy introduced as it provided a minor gain in the
load time of the elements, pre-scheduler.
For the base-files subset of the debian-stack, this gain was ~2 seconds.
However, with it, came a fairly significant cost to peak memory usage (~100M).
Additionally, the benchmarks have highlighted that this patch has introduced
some cost during build time, where the build of the base-files subset of the
debian-like project has increased by ~90 seconds and peak memory by ~100M
This reverts commit 746aa7a63c4d06b1caa012a35c970b43e0166faa and
commit d55039c33685d267fd8834ecc5d16030c1385325.
|
|\
| |
| |
| |
| |
| |
| | |
setup.py: Include buildstream.testing datafiles in package_data
Closes #1008
See merge request BuildStream/buildstream!1309
|
|/
|
|
|
|
|
| |
* Remove `include_package_data=True` from setup.py
* Collect datafiles for buildstream.testing by adding them to `package_data`
Fixes #1008
|
|\
| |
| |
| |
| |
| |
| | |
stack.py: Stop creating empty bst directory
Closes #998
See merge request BuildStream/buildstream!1301
|
|/
|
|
|
|
|
|
|
|
|
| |
It seems like this hack was added as a workaround for OSTree's
limitations to commit an empty directory, originally in
https://gitlab.com/BuildStream/buildstream/commit/91940a6cba193ac0e20c01008335617847be27a5.
Since we do not have this limitation anymore, stop creating this empty
directory.
Fixes #998.
|
|\
| |
| |
| |
| | |
Fix building the docs
See merge request BuildStream/buildstream!1303
|