* test: remove target validator test
these appear to have too much maintenance burden as they only ever
add more over time
* fix: allow google_apis_ps16k and google_apis_playstore_ps16k as valid targets
Fixes#403
Co-authored-by: Yang <reactivecircus@gmail.com>
* fix: remove target validation entirely, any valid sdkmanager target will work
no longer requires code changes here to access new targets
---------
Co-authored-by: Yang <reactivecircus@gmail.com>
* Optimize config.ini updates by batching shell executions
- Reduce up to 5 separate shell executions to 1 for AVD configuration
- Improves performance by eliminating redundant process spawns
- Add comprehensive efficiency report documenting all identified improvements
- All existing tests pass, no functional changes
Co-Authored-By: Yang <reactivecircus@gmail.com>
* Add compiled JS files and remove standalone efficiency report
- Include built lib/emulator-manager.js with batched config optimization
- Remove EFFICIENCY_REPORT.md as requested
- Efficiency report content will be moved to PR description
Co-Authored-By: Yang <reactivecircus@gmail.com>
---------
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: Yang <reactivecircus@gmail.com>
* Avoid changing the owner of unnecessary files
Source files shouldn't matter for the purposes of this tool.
% ls -R $ANDROID_HOME/sources/android-34 | wc -l
17750
* Remove non-existant directory
* Avoid another unnecessary chown
* Push js
* Add port parameter
* Fix a typo in test description
* Fix avd not being started with correct port
* Fix wrong port being used to kill an emulator if there was an exception
Previous code worked in a wrong way.
1. Getting the first char of a build number is mistake because now numbers start with 1, but whole number higher than the number which starts from 7.
2. URL was incorrect for arm mac machines. It needed to add `aarch64` suffix
* Support non-integer API level
* update lib
* Add test
* No need to install platforms.
Mostly application will use different SDK platform.
* Update lib
* Add `pre-emulator-launch-script` to `action.yml`
* Add `pre-emulator-launch-script` in `main.yml`
* Implement pre emulator launch script
* Try to use `pushd`
* print working directory
* Add `README`
* Run `npm run build`
* Fix `working-directory`
* Run `npm run build`
* Run pre emulator launch in group
* Run `npm run build`
* Update README.md
Co-authored-by: Yang <reactivecircus@gmail.com>
* Add type "pre-emulator-launch-script"
Co-authored-by: Yang <reactivecircus@gmail.com>
* Allow increasing disk size - Issue #184
* Fix checkDiskSize validator
* Fix lint errors
* Size modifier without a number is unacceptable
"G" is not the same as "1G"
* Implement checkDiskSize validator tests
* Update build-tools to 31.0.0. Add API 31 emulator to workflow. Update SDK command-line tools to 5.0
* Update test fixture dependencies.
* Test API 31 with macos-11.
* Add android:exported to AndroidManifest.
* Format
* Add non-mobile targets to allowlist
* Add new targets to input validator tests
* Update built input-validator.js
* Update README with new targets
* Update action.yml with new targets
The exec() command only runs a program with literal argument strings.
It does not know how to do things like expanding environment variables,
or piping/redirection. Hence the way to get "shell intelligence"
is to use exec() to run the `sh` process with the command line string
as a parameter.
But the method being used to do this was:
exec.exec(`sh -c \\"${script}"`)
This has problems, because wrapping the script in quotes creates
issues when the script itself contains quotes. Escaping the string
correctly is a non-trivial problem, which also would create "noise"
in the debug output which would add confusion.
http://mywiki.wooledge.org/BashFAQ/050
The easiest way to work around this here is to use the array form of
exec() to pass exactly two parameters to `sh`. No manipulation of the
script string is needed with this approach:
exec.exec("sh", ["-c", script])
There are more cases in the project of `sh -c` usage which should also
be changed, and likely abstracted (shellExec()?) But this small patch
just fixes the most important case for the user-provided script.
---
Additionally, this removes the escaped backslash (`\\`) from the
start of the executed command. That was presumably to suppress the
use of aliases:
https://unix.stackexchange.com/questions/524254/why-are-backslashes-included-in-this-shell-script
But because this is invoking a non-interactive shell session, aliases
would not apply. Hence the extra backslash shouldn't be needed.
Mike Hardy