Skip to content

Conversation

@pull
Copy link

@pull pull bot commented May 20, 2023

See Commits and Changes for more details.


Created by pull[bot] (v2.0.0-alpha.4)

Can you help keep this open source service alive? 💖 Please sponsor : )

@pull pull bot added ⤵️ pull merge-conflict Resolve conflicts manually labels May 20, 2023
@pull pull bot requested a review from cpsauer May 20, 2023 02:12
LeFrosch and others added 27 commits June 13, 2025 11:50
Do not mark the execution root as dirty before calling refresh for clwb

Followup for after further investigation #7673
Results of the investigation:
- Refresh after sync is required to pick up changes from the exec root
- Found an unnecessary refresh, only refreshing the execution root should be sufficient
- Mark dirty is not required (however, might cause a race condition with the notifier)
The truth8 extension as been bundled into the standard truth package and is therefore no longer required as an additional dependency. And somehow rules_kotlin 2.1.7 did not like truth8.

Also removes the codeql patches.
The BuildResultHelper abstraction is no longer used post-cleanup, and
will be deleted. The GetArtifactsException that currently is inside the
interface is used, so this change extracts it out into a separate class.

Bug:374906681

Test: existing tests
Change-Id: Ic89a342707a3b1b562fcfcc03f2260956a726eb8

AOSP: 1b413f38e00ac9d4533b66012cec51bd81ff700b

Co-authored-by: Googler <[email protected]>
Adds simple heuristic whether clang-cl is used by looking at the current compiler name provided by Bazel.
LeFrosch and others added 30 commits December 9, 2025 12:31
This PR adds support for warding CLion's proxy settings to Bazel. However, this is disabled by default and can be enabled in the settings: Bazel Settings | Forward proxy settings
…8029)

* Rename .java to .kt

* Prevent freezes in AspectSyncSourceToTargetMap.getSourceToTargetMap

getSourceToTargetMap might be called under read action and it's
a quite heavy operation. we can bring cancellation checks into it,
but

a. it might be called under non-cancellable read action
b. it might take tens of seconds to compute, so no readaction
can survive this long

so if we're called under readaction, we just return null and schedule
background computation, and hope we will be called some time later.
but we expect to be called from ProjectTargetManagerImpl on the
background thread, so we can build map, but in case we're invoked
from someone else before, background computation will prevent the
freeze
Use of fdebug-compilation-dir in macos debug runs
leads to 100% remote cache miss because the path for
compilation dir is valid for the current user only.
The change reworks the original PR #4403 and provides a
more correct fix for #2101 by launching lldb in the
proper working directory and using relative paths to set
breakpoints
this turned out to be ok to have these files
as nonexisting because rulesets might point to
additional locations which do not contain files
we're looking for
After #8029 run test gutters almost always display "Nothing here"
Now SyncCache checks for cancellations and `ReverseDependencyMap.get`
is considered as canceled, because `ProgressManager.checkCancelled`
ensures that Job is active, but it's not because of the following:

`PendingAsyncTestContext.fromTargetFuture` used directExecutor,
which might reuse / reuses future thread, but future is executed in
coroutine scope converted to future, so scope's Job is completed when
future is done. We should avoid using directExecutor as mentioned in
Futures documentation
* use lldb from installer

since 2025.3 CLion prefers to use lldbfrontend from
native-debug-plugin and soon lldbfrontend will be removed
from the CLion installer. Unfortunately, our tests are run
with the flat classpath and we can't rely on built-in code
do detect lldbfrontend based on some NDP classloader.

let's rely on the flag to use installer-provided lldbfrontend
and decide on the next steps later one it's deprecated

* Disable soft_fail for clion-oss-under-dev

All the issues are addressed and we can have soft_fail as false now
This was spotted in #8046 discussion, that there might be non-hermetic
libraries used with debug information available, so reducing the scope
for relativization
We found that providing fallback configuration to the backend leads to recursive watch roots being set up for include paths mentioned in the fallback config. We know this has very negative performance implications because of greediness in VfsWarmQueue. This extension point allows us to disable fallback configurations for Bazel projects.

However, since the extension point was introduced in 2025.2.5, it is not save to simply instantiate the implementation class and we need to do some manual registration magic.
Since clang-cl is interface compatible with MSVC we can simply use the MSVC switch builder and avoid some build issues caused by the /clang: prefix.
Previously the execution headless tests only run a catch2 target. This PR adds a new target for gtest and also runs both targets using the Debug executor.
Currently we only support the copts rule attribute. However, it is also possible to set language specific options with cxxopts and conlyopts rule attributes. This PR adds support for these attributes.
…is disabled (#8066)

Extra quotes injected into a source filename cause LLDB breakpoints to stop working. Use toSystemIndependentName to address issue with this option passed down to cmake with rules_foreign_cc as CMake requires / in paths.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

⤵️ pull merge-conflict Resolve conflicts manually

Projects

None yet

Development

Successfully merging this pull request may close these issues.

10 participants