Autoroll from main to 27.lts - #884
Open
cobalt-github-releaser-bot wants to merge 10 commits into
Open
cobalt-github-releaser-bot wants to merge 10 commits into
cobalt-github-releaser-bot wants to merge 10 commits into
Conversation
Refer to original PR: #11234 This disables the pre-existing `enable_devtools_frontend` buildflag on Cobalt builds and conditionally removes `devtools_resources.pak` from `//cobalt/shell:pak`, plus some other changes to ensure non-Android builds succeed. Remote devtools are still functional. `enable_devtools_frontend` is already disabled upstream on Android, so there should be no impact there. For 3P builds, this reduces the size of `cobalt_shell.pak` from 5.3 MB to 2.0 MB. Bug: 525402992 Bug: 523267920 (cherry picked from commit f5e2be9)
Refer to original PR: #10951 This PR adds the `cobalt-rdk-deploy` agent skill to the repository under `.agent/skills/`. The skill provides guidelines and execution protocols for Gemini/Jetski to build, deploy, run, and debug Cobalt on RDK devices using the `deploy_rdk.py` script. ### Guidelines included: 1. **Rely on help:** Always run `--help` first to discover available execution modes and flags. 2. **Execute via script:** Formulate and run the `deploy_rdk.py` script directly to interact with the device. 3. **Avoid manual device interaction:** Explicitly forbids running raw shell commands (via `adb shell` or `ssh`) directly on the target device for pre-flight checks (trusting the script to handle them). 4. **Toolchain assumptions:** Assume the compilation toolchain is present by default and do not run `--setup-toolchain` or perform manual pre-emptive checks (like checking `$RDK_HOME`) unless compilation fails. 5. **No-RBE default:** Avoid passing `--no-rbe` by default (only use it if compilation fails). Bug: 502702565 TAG=agy CONV=e3d4c794-091a-4ae7-8cd1-810eebeb07ee (cherry picked from commit aa9aef6)
….pak Refer to original PR: #11291 This saves 700 KB by removing the resources required for the chrome://tracing page, which can't be accessed on Cobalt. The tracing backend is untouched. Bug: 523267920 (cherry picked from commit 939e473)
Refer to original PR: #11304 This not only adds functionality to support a second to staging autoroll, but also adds functionality to support a non-main source branch. To handle various .gitmodules issues such as corruption or conflicts, the logic in resolve_conflicts was updated. Due to the larger and more complex Chromium source cherry-picks, the checkouts can no longer be sparse / none blobs. Bug: 522987128 (cherry picked from commit b7ec7db)
Refer to original PR: #11317 The Crashpad handler executable is now expected to be under a "native_target" parent directory, which is itself alongside the loader app executable. This was originally thought of as a "hack" but we have sinced aligned on the decision to promote this path as our new stanard (see #9841). Issue: 476394156 (cherry picked from commit a67bc99)
…4 Android builds Refer to original PR: #11296 This change limits the compilation of the validating command decoder to 64-bit Android builds (target_cpu == "arm64"). Since Nvidia Shield is the only Android TV device that runs on an arm64 chip and requires the validating command decoder as a fallback (due to driver bugs under the passthrough decoder), keeping it on arm64 ensures Shield remains functional. Disabling it on 32-bit arm builds saves ~320 KB of binary size on all other memory-constrained Android TVs. Bug: 521911827 TAG=agy CONV=821cb3d0-a384-4064-bf2a-e82940e74df9 (cherry picked from commit 92b3f5d)
…headers Refer to original PR: #11290 This fixes an issue with diagrams in codeblocks expanding to include surrounding text in the `starboard/event.h` file. Additionally there were some deletions to raspberry pi docs so regenerating for that too. Bug: 529822255 (cherry picked from commit b3d03bd)
Refer to original PR: #11287 Turn SB_DCHECK_GE to SB_CHECK_GE in MediaCodecDecoder when receiving input and output buffers from Android MediaCodec. This ensures that invalid negative indices are caught immediately in all builds. Bug: 329686979 (cherry picked from commit 73e5cf5)
Refer to original PR: #11318 Due to AOSP need, StarboardBrdige was renamed. This PR reverts the rename and instead put shared code in a new base class that StarboardBridge inherits from. Bug: 533443758 (cherry picked from commit fb366eb)
…ache Refer to original PR: #11248 DRAFT This PR introduces the EnableOptimizedV8CodeCache switch. When enabled: 1. Increases the V8 Generated Code Cache quota to 5 MB in `CobaltContentBrowserClient::GetGeneratedCodeCacheSettings`. 2. Filters out small (< 1 KB) bytecode snippets in `GeneratedCodeCache::WriteEntry` to avoid LRU table clutter. YouTube Living Room's compiled V8 script payload (kabuki.base, kabuki.main, and tv-player-es6.js) sums to ~3.96 MB. Under the default 3 MB V8 code cache quota, core scripts continuously evict each other across warm starts. Locally on ADT-4, seeing reducing warm startup time to video player ready by ~855 ms (from 2,477 ms to 1,622 ms). Bug: 532266766 --------- Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> (cherry picked from commit e91d641)
cobalt-github-releaser-bot
requested review from
briantting,
linxinan-yt and
oxve
September 8, 2026 23:15
Contributor
🤖 Gemini Suggested Commit Message💡 Pro Tips for a Better Commit Message:
|
![gemini-code-assist[bot]](/web-api/iframe-proxy?url=https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Code Review
This pull request refactors the Android Starboard bridge by introducing BaseStarboardBridge and simplifying StarboardBridge, disables the DevTools frontend for Cobalt to optimize binary size, restricts the validating command decoder to 64-bit Android builds, and introduces an optimized V8 code cache option. Feedback has been provided regarding a potential AttributeError in the LTS autoroll script when resolving submodule conflicts, and an unsafe publication of this in the BaseStarboardBridge constructor before its final fields are fully initialized.
Comment on lines
+140
to
+146
| match = re.search(r'160000 ([a-f0-9]+) 3', ls_files_out) | ||
| theirs_sha = match.group(1) | ||
| run([ | ||
| 'git', 'update-index', '--add', '--cacheinfo', | ||
| f'160000,{theirs_sha},{path}' | ||
| ]) | ||
| unmerged_files.pop(path, None) |
Contributor
There was a problem hiding this comment.
In a modify/delete conflict on a submodule, the stage 3 ('theirs') entry will not exist in the git index. In this scenario, re.search will return None, and calling match.group(1) will raise an AttributeError, crashing the autoroll script. We should check if match is not None before extracting the SHA.
Suggested change
| match = re.search(r'160000 ([a-f0-9]+) 3', ls_files_out) | |
| theirs_sha = match.group(1) | |
| run([ | |
| 'git', 'update-index', '--add', '--cacheinfo', | |
| f'160000,{theirs_sha},{path}' | |
| ]) | |
| unmerged_files.pop(path, None) | |
| match = re.search(r'160000 ([a-f0-9]+) 3', ls_files_out) | |
| if match: | |
| theirs_sha = match.group(1) | |
| run([ | |
| 'git', 'update-index', '--add', '--cacheinfo', | |
| f'160000,{theirs_sha},{path}' | |
| ]) | |
| unmerged_files.pop(path, None) | |
| else: | |
| log(f'Failed to find stage 3 SHA for submodule {path}') |
Comment on lines
+122
to
+129
| // Make sure the JNI stack is properly initialized first as there is a | ||
| // race condition as soon as any of the following objects creates a new thread. | ||
| BaseStarboardBridgeJni.get().initJNI(this); | ||
|
|
||
| mAppContext = appContext; | ||
| mActivityHolder = activityHolder; | ||
| mServiceHolder = serviceHolder; | ||
| mArgs = args; |
Contributor
There was a problem hiding this comment.
Initializing the final fields (mAppContext, mActivityHolder, mServiceHolder, mArgs) after passing this to BaseStarboardBridgeJni.get().initJNI(this) exposes an incompletely constructed object to native code. If the native code synchronously invokes any @CalledByNative methods or spawns a thread that accesses these fields, it can result in a NullPointerException or memory visibility issues under the Java Memory Model. These simple reference assignments should be moved to the top of the constructor before initJNI(this).
Suggested change
| // Make sure the JNI stack is properly initialized first as there is a | |
| // race condition as soon as any of the following objects creates a new thread. | |
| BaseStarboardBridgeJni.get().initJNI(this); | |
| mAppContext = appContext; | |
| mActivityHolder = activityHolder; | |
| mServiceHolder = serviceHolder; | |
| mArgs = args; | |
| mAppContext = appContext; | |
| mActivityHolder = activityHolder; | |
| mServiceHolder = serviceHolder; | |
| mArgs = args; | |
| // Make sure the JNI stack is properly initialized first as there is a | |
| // race condition as soon as any of the following objects creates a new thread. | |
| BaseStarboardBridgeJni.get().initJNI(this); |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
9 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Automated cherry-pick roll to 27.lts.
Original pull requests:
To merge these changes, run the autoroll script: