| Commit message (Collapse) | Author | Age |
|---|
| |
|
|
|
|
|
|
|
|
| |
*** Reason for rollback ***
Fixed duplicate derived inputs bug. Test is in diffbase.
RELNOTES[INC]: If the same artifact is generated by two distinct but identical actions, and a downstream action has both those actions' outputs in its inputs, the artifact will now appear twice in the downstream action's inputs. If this causes problems in Skylark actions, you can use the uniquify=True argument in Args.add_args.
PiperOrigin-RevId: 205863806
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
*** Reason for rollback ***
Still bugs lurking. See linked bug.
*** Original change description ***
Automated rollback of commit eb587075b0d6ffab1cf9e69ede1b7e547905e547.
*** Reason for rollback ***
Depot has been fixed.
RELNOTES[INC]: If the same artifact is generated by two distinct but identical actions, and a downstream action has both those actions' outputs in its inputs, the artifact will now appear twice in the downstream action's inputs. If this causes problems in Skylark actions, you can use the uniquify=True argument in Args.add_args.
PiperOrigin-RevId: 204997569
|
| |
|
|
|
|
|
|
|
|
| |
*** Reason for rollback ***
Depot has been fixed.
RELNOTES[INC]: If the same artifact is generated by two distinct but identical actions, and a downstream action has both those actions' outputs in its inputs, the artifact will now appear twice in the downstream action's inputs. If this causes problems in Skylark actions, you can use the uniquify=True argument in Args.add_args.
PiperOrigin-RevId: 204827477
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
than the graph version when that is feasible.
* It's not feasible when the computation accesses outside state, i.e. is non-hermetic, so see below.
* It's also more complicated (and not worth the trouble) when the computation is taking place just for the error status.
Have SkyFunctionName declare whether the function it corresponds to is hermetic or non-hermetic. Only non-hermetically-generated SkyValues can be directly marked changed, and non-hermetic SkyFunctions have their values saved at the graph version, not the max of the child versions. All SkyFunctions are hermetic except for the ones that can be explicitly dirtied.
A marked-hermetic SkyFunction that has a transient error due to filesystem access can be re-evaluated and get the correct version: if it throws an IOException at version 1 and then, when re-evaluated at version 2 with unchanged dependencies, has a value, the version will be version 1.
All Skyframe unit tests that were doing non-hermetic things to nodes need to declare that those nodes are non-hermetic. I tried to make the minimal set of changes there, so that we had good incidental coverage of hermetic+non-hermetic nodes. Also did some drive-by clean-ups around that code.
Artifacts are a weird case, since they're doing untracked filesystem access (for source directories). Using max(child versions) for them gives rise to the following correctness bug: 1. do a build at v1 that creates a FileStateValue for dir/ at v1. Then at v2, add a file to dir/ and do a build that consumes dir/ as a source artifact. Now the artifact for dir/ will (incorrectly) have v1. Then at v1, do that build again. We'll consume the "artifact from the future". However, this can only have an effect when using the local action cache, since the incorrect value of the artifact (the mtime) is only consumed by the action cache. Bazel is already broken in this way (incremental builds don't invalidate directories), so this change doesn't make things worse.
PiperOrigin-RevId: 204210719
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
*** Reason for rollback ***
Breaks Javascript compilation.
There's currently no way to iterate over a depset of Artifacts and deduplicate by identical paths in Skylark. This means that actions that want to do something once per Artifact in a depset (add a flag to the command line with the path of the Artifact for instance) will have duplicate entries if there are multiple Artifacts with the same path.
This is not a true automated rollback, since I tried to make this as minimal as possible, to avoid merge conflicts and keep some of the benefits of the rolled back CL. In particular, the tests that were changed in the original CL to give artifacts their correct owners did not need to be changed back, since the owners are still correct.
Moreover, this effectively contains rollbacks of unknown commit and https://github.com/bazelbuild/bazel/commit/39d6f89644107a8f7c080c4f4aec8527c1a73954, but keeps test coverage from those CLs as well. Comments added to CL thread where not a clean rollback (there are plenty of files not changed at all: ActionArtifactCycleReporter is the main wart, since it can still handle SkyKeys with Artifact as the argument, but Artifact is no longer the argument to any SkyKey).
RELNOTES: None
*** Original change description ***
Make Artifact#equals take the owner into account for derived artifacts.
Derived artifacts' owners are important because they are used to determine the artifact's generating action. Source artifacts' owners are not used in this way, so I left them alone.
This allows us to get rid of most uses of ArtifactSkyKey. We may be able to delete it entirely in a follow-up.
PiperOrigin-RevId: 201464780
|
| |
|
|
|
|
|
| |
The memory savings from this flag are not worth the complexity, and it interferes with action restarting.
RELNOTES: Remove support for --discard_actions_after_execution.
PiperOrigin-RevId: 201077905
|
| |
|
|
|
|
|
|
| |
Derived artifacts' owners are important because they are used to determine the artifact's generating action. Source artifacts' owners are not used in this way, so I left them alone.
This allows us to get rid of most uses of ArtifactSkyKey. We may be able to delete it entirely in a follow-up.
PiperOrigin-RevId: 199836436
|
| |
|
|
|
|
| |
(minor) ActionFS now implements MetadataProvider.getInput
PiperOrigin-RevId: 199575194
|
| |
|
|
|
|
|
|
|
| |
Given a target (for example from a skylark aspect), one will be able to access a list of actions that the target generated using "target.actions". This is without additional memory footprint.
Actions themselves are not fully exposed to skylark (and thus there isn't much meaning to gather from them in skylark yet). Access methods will follow soon.
RELNOTES: None.
PiperOrigin-RevId: 188098079
|
| |
|
|
|
|
| |
come from Skylark, so we shouldn't crash. As a safety measure, subclasses of ActionLookupValue are now responsible for detecting action conflicts themselves.
PiperOrigin-RevId: 187095271
|
| |
|
|
|
|
| |
mechanism as for normal actions, have the ActionTemplateExpansionFunction look the template up when needed.
PiperOrigin-RevId: 185861672
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Path and PathFragment have been replaced with String-based implementations. They are pretty similar, but each method is dissimilar enough that I did not feel sharing code was appropriate.
A summary of changes:
PATH
====
* Subsumes LocalPath (deleted, its tests repurposed)
* Use a simple string to back Path
* Path instances are no longer interned; Reference equality will no longer work
* Always normalized (same as before)
* Some operations will now be slower, like instance compares (which were previously just a reference check)
* Multiple identical paths will now consume more memory since they are not interned
PATH FRAGMENT
=============
* Use a simple string to back PathFragment
* No more segment arrays with interned strings
* Always normalized
* Remove isNormalized
* Replace some isNormalizied uses with containsUpLevelReferences() to check if path fragments try to escape their scope
* To check if user input is normalized, supply static methods on PathFragment to validate the string before constructing a PathFragment
* Because PathFragments are always normalized, we have to replace checks for literal "." from PathFragment#getPathString to PathFragment#getSafePathString. The latter returns "." for the empty string.
* The previous implementation supported efficient segment semantics (segment count, iterating over segments). This is now expensive since we do longer have a segment array.
ARTIFACT
========
* Remove Path instance. It is instead dynamically constructed on request. This is necessary to avoid this CL becoming a memory regression.
RELNOTES: None
PiperOrigin-RevId: 185062932
|
| |
|
|
|
|
|
|
| |
essentially promote OwnedArtifact to ArtifactSkyKey and rename it to ArtifactSkyKey. The king is dead...
Also add some other execution-phase codecs.
PiperOrigin-RevId: 184552706
|
| |
|
|
|
|
| |
SpecialArtifact.
PiperOrigin-RevId: 184539696
|
| |
|
|
|
|
| |
owner and fix up BuildConfigurationValue.Key. ConfiguredTargetKey is going to need some modifications to AutoCodec: probably the long-awaited static "create" method.
PiperOrigin-RevId: 182630181
|
| |
|
|
|
|
|
|
|
|
|
| |
This class represents a root (such as a package path or an output root) used for file lookups and artifacts. It is meant to be as opaque as possible in order to hide the user's environment from sky keys and sky functions.
Roots are used by RootedPaths and ArtifactRoots.
This CL attempts to make the minimum number of modifications necessary to change RootedPath and ArtifactRoot to use these fields. Deprecated methods and invasive accessors are permitted to minimise the risk of any observable changes.
RELNOTES: None
PiperOrigin-RevId: 182271759
|
| |
|
|
|
|
| |
This is slightly more descriptive, and we will potentially want to use the name Root for a broader concept shared between ArtifactRoot and RootedPath.
PiperOrigin-RevId: 182082367
|
| |
|
|
|
|
| |
removing the layer of indirection of getting SkyKey out of ActionLookupKey, which uses more memory for no reason.
PiperOrigin-RevId: 181658615
|
| |
|
|
|
|
|
| |
This key context can be used by actions to share partial key computations, for instance when computing MD5s for nested sets.
RELNOTES: None
PiperOrigin-RevId: 177359607
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
instead of assuming BUILD.
- Default the list to the same value as PackageLookupFunction:
BUILD.bazel, BUILD.
- Move BuildFileNames to the packages package, so it is more generally
available.
Part of #4056.
Change-Id: Ie12512b492cd7d47a9e56ec3bc209f829feaf4b5
PiperOrigin-RevId: 177261295
|
| |
|
|
|
|
|
|
| |
Blaze had its own class to avoid GC from varargs array creation for the precondition happy path. Guava now (mostly) implements these, making it unnecessary to maintain our own.
This change was almost entirely automated by search-and-replace. A few BUILD files needed fixing up since I removed an export of preconditions from lib:util, which was all done by add_deps. There was one incorrect usage of Preconditions that was caught by error prone (which checks Guava's version of Preconditions) that I had to change manually.
PiperOrigin-RevId: 175033526
|
| |
|
|
|
|
| |
or no edges. Also add option to disable checks in MemoizingEvaluatorTest that don't make sense for implementations that don't keep track of dirty nodes. Also extract RecordingDifferencer to an interface. And add a test for the situation that a node changes during a build that it's not requested, and which fails, necessitating cleanup.
PiperOrigin-RevId: 171616817
|
| |
|
|
|
|
|
|
|
|
|
|
| |
'create' method.
This paves the way for changing PathFragment to e.g. an abstract class with multiple subclasses. This way we can split out the windows-specific stuff into one of these concrete classes, making the code more readable and also saving memory (since the shallow heap size of the NonWindowsPathFragment subclass will hopefully be smaller than that of the current PathFragment).
This also lets us pursue gc churn optimizations. We can now do interning in PathFragment#create and can also get rid of unnecessary intermediate PathFragment allocations.
RELNOTES: None
PiperOrigin-RevId: 152145768
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
--noexperimental_enable_critical_path_profiling flags are all specified, then Bazel will delete Actions from ActionLookupValues as they are executed in order to save memory.
Because an already-run action may output an artifact that is only requested later in the build, we need to maintain a way for the artifact to look up the action. But in most cases we don't need to keep the action itself, just its output metadata.
Some actions unfortunately are needed post-execution, and so we special-case them.
Also includes dependency change with description:
Move action out of key. This keeps action references from polluting the graph -- actions are just stored in one SkyValue, instead of being present in SkyKeys.
This does mean additional memory used: we have a separate ActionLookupData object per Action executed. That may reach ~24M for million-action builds.
PiperOrigin-RevId: 151756383
|
| |
|
|
|
|
|
| |
digest are mutually exclusive.
--
MOS_MIGRATED_REVID=128843642
|
|
|
1. Adds ActionTemplateExpansion{Function, Value} for ActionTemplate expansion.
2. Changes ArtifactFunction to support evaluating TreeFileArtifacts and TreeArtifacts generated by ActionTemplate.
--
MOS_MIGRATED_REVID=124160939
|
janakr
tomlu
shahan
cparsons
cpeyser
John Cater
nharmata
Rumou Duan