After upgrading `react-native` to `0.72.5` we frequently started seeing the _red screen of death_ on both `Android` and `iOS` simulators right after the app was built and installed.
This used to happen because our workflow required us to do the following :
- `make run-clojure`
- `make run-metro`
- `make run-ios` OR `make run-android`
The problem with this approach was after `metro` was started the `iOS`, `Android` build step would change the files that `metro` couldn't handle and hence metro would go out of sync.
The quick fix back then was to restart `metro` terminal and to open the app again from the simulator.
This was however not a good DX.
This commit fixes that.
We no longer rely on `react-native` cli to generate and deploy debug builds on simulators. We take control of the process via our own script. The new workflow introduced in this commit will first build the app, then install the app on the simulators and then start metro terminal. When `metro` is successfully running the script will then open the app.
The new workflow now is :
- `make run-clojure`
- `make run-ios` OR `make run-android`
fixes#18493
We enabled `hermes` for android in the `react-native` upgrade to `0.72.5`
Although things seemed fine but developers were seeing frequent crashes in their local environment.
After some investigation the crashes were traced to max native call stack depth in `hermes` engine.
Disabling `hermes` for local debug builds helps fix that issue.
This commit disables `hermes` by default with the help of a exporting an environment variable in the `make run-android` command.
It is annoying that this also modifies `android/gradle.properties` so we keep `hermesEnabled` as `false` there as well.
We also enable `hermes` when generating release builds so that we can take advantage of `hermes` engine in release builds.
We also add a log to print whether `hermes` is enabled or not. I think its helpful to have this so that we know whether `hermes` is enabled or not.
Fixes partially #15595
In order to build less targets, when not needed we introduce this
mapping logic.
If only specific ABI is required - status-go will have the same
arhitecuture.
For some unknown to me reason we are using a different Yarn call to
Shadow-cljs to generate the JSBundle for iOS builds, while the one
created by the Android derivation shoudl be exactly the same.
I'm changing the target to just be `make jsbundle` while keeping aliases
referencing old naming, and moving things around in `nix` folder to
reflect the fact that the derivation is no longer Android-specific.
Also, crucially, I've changed the `import` in `index.js` to use the
`./result/index.js` path, since that's what Nix creates. I'm not sure if
this clashes with any developer workflow that takes place locally, so
I'd appreciate some testing from developers.
Depends on: https://github.com/status-im/status-jenkins-lib/pull/67
Signed-off-by: Jakub Sokołowski <jakub@status.im>
Notable upgrades:
* Go `1.17.11` to `1.18.6`
* NodeJS `16.15.0` to `16.17.1`
* Clojure `1.11.1.1139` to `1.11.1.1165`
* Ruby Gem `3.2.26` to `3.3.20`
* Bundler `2.3.9` to `2.3.22`
* Git `2.36.1` to `2.37.3`
* Curl `7.83.1` to `7.85.0`
* OpenSSL `1.1.1o` to `3.0.5`
* PatchELF `0.14.5` to `0.15.0`
* Android SDK Platform Tools `33.0.1` to `33.0.2`
Signed-off-by: Jakub Sokołowski <jakub@status.im>
This passing of Watchman socket was implemented in order to avoid this:
```
Error: EMFILE: too many open files, watch
at FSEvent.FSWatcher._handle.onchange (node:internal/fs/watchers:204:21)
Emitted 'error' event on NodeWatcher instance at:
at NodeWatcher.checkedEmitError (/private/tmp/nix-build-status-mobile-build-nightly-android.drv-0/node_modules/sane/src/node_watcher.js:143:12)
at FSWatcher.emit (node:events:527:28)
at FSEvent.FSWatcher._handle.onchange (node:internal/fs/watchers:210:12) {
errno: -24,
syscall: 'watch',
code: 'EMFILE',
filename: null
}
```
Which is caused by `jest-haste-map` used by `metro` starting to watch
the filesystem for file changes, which is pointless when doing a
one-off build using Nix.
But by setting `CI=true` we can make `metro` not start this waching of
files in the first place, removing the need for use of Watchman entirely.
By entirely dropping use of Watchman we also fix the following issue:
```
[cli] unable to talk to your watchman on /tmp/tmp-status-mobile-ABC/jenkins-state/sock! (Permission denied)
```
Which happens on multi-user Nix installations becuase the user that the
Nix build is executed as is not the same as the user that starts
Watchman and creates the socket file.
Issue: https://github.com/status-im/status-mobile/issues/13783
Signed-off-by: Jakub Sokołowski <jakub@status.im>
This way the name of the repo makes at least some sense and
matches the `status-desktop` repo naming.
Also updated `status-jenkins-lib` since it also contained
references to `status-react` repo and job names.
Signed-off-by: Jakub Sokołowski <jakub@status.im>
This has several benefits:
* Less abuse of `extra-sandbox-paths` Nix option
* Less inputs to the Android release build derivation
* Easier for users to sign the build themselves
* Simplification of `scripts/release-android.sh`
* Preparation for building using Nix Flakes
The only two remaining credentials passed via `extra-sandbox-paths` is
the Infura and OpenSea API keys, and there is no way around that other
than passing them via Nix arguments, but that would cause them to end up
in `/nix/store` as part of `.drv` files.
I'm also renaming `release-fdroid` to `build-fdroid` to be consistent.
Depends on: https://github.com/status-im/status-jenkins-lib/pull/42
Signed-off-by: Jakub Sokołowski <jakub@status.im>
Incorrect reference to `nimbus.src-override` causes Gradle builds to fail with:
```
Caused by: org.gradle.internal.resolve.ModuleVersionResolveException: Could not resolve status-im:status-go:e940434.
```
Due to this `if` clause not triggering:
a573ae70e5/modules/react-native-status/android/build.gradle (L4-L8)
Signed-off-by: Jakub Sokołowski <jakub@status.im>
This will make identifying installed software easier.
It can also be used to generate F-Droid RPs more easily.
Depends on:
https://github.com/status-im/status-jenkins-lib/pull/35
Signed-off-by: Jakub Sokołowski <jakub@status.im>
The env variable `INFURA_TOKEN` is used at build time of JS bundle, not
the final APK file. We never passed the `secretsFile` to the
derivation for JS bundle so it never saw the `INFURA_TOKEN`.
Signed-off-by: Jakub Sokołowski <jakub@status.im>
This updates a bunch of things in the Jenkins logic.
For details of that see:
https://github.com/status-im/status-react-jenkins/pull/12
It grew out of the changes I made to make PR builds available in Desktop client:
https://github.com/status-im/nim-status-client/pull/507
Changes:
- Bump `status-react-jenkins` library to newer version, probably `1.2.0`
- Update use of `utils.pkgFilename()` in `ci/Jenkinsfile.combined`
- Make all iOS builds go into the same folder: `status-ios`
- Drop e2e build type, infer it from Android ABIs set to `x86` only
- Simplify logic in `nix/mobile/android/release.nix.`
Signed-off-by: Jakub Sokołowski <jakub@status.im>
Changes:
- Simplified `default.nix` and `shell.nix`
- Moved the default shell to `nix/shell.nix`
- Dropped unnecessary merge from `nix/shells.nix`
- Fixed `nix/lib/getConfig.nix` to return default on `null`
- Expanded `nix/DETAILS.md` with more info
- Added links to presentations in `nix/README.md`
- Fixed a few typos
Signed-off-by: Jakub Sokołowski <jakub@status.im>
Changes:
* Create `nix/config.nix` with `config` defaults
* Add `nix/tools/gradlePropParser.nix` for reading `gradle.properties`
* Add `nix/mobile/android/keystore.nix` for generating a keystore
* Load keystore generation in `nix/mobile/android/default.nix`
* Use generated keystore if it's not provided via `config`
* Add `-deststoretype pkcs12` in `scripts/generate-keystore.sh`
* Add `nix/lib/assertEnvVarSet.nix` for checking if env var is set
Signed-off-by: Jakub Sokołowski <jakub@status.im>
Before in order to create `deps.nodejs-patched` the `deps.gradle` would
also have to be downloaded in order to patch `build.gradle` files with
path to the Gradle dependencies in Nix store.
It turns out just replacting lines referencing `mavenCentral()`,
`google()`, and `jcenter()` in `repositories` block is enough to make
Gradle properly fetch dependencies from repo provided via the command
line `-Dmaven.repo.local='${deps.gradle}` option.
This should reduce the required size for shells that don't use Gradle.
Signed-off-by: Jakub Sokołowski <jakub@status.im>
Becuase `clojure` shell used unpatched `node_modules` while the
`android` shell used the patched version when starting these shells
the `node_modules` would keep being switched between them wasting time.
Changes:
- Move `nix/tools/patchNodeModules.nix` to `nix/deps/nodejs-patched`
- Make `nix/deps/nodejs-patched` a normal derivation without extra arguments
- Use `deps.nodejs-patched` in `nix/mobile/android` shell
- Use `deps.nodejs-patched` in `nix/shells.nix` node shell
- Use `with pkgs` to reduce arguments in `nix/mobile/android/release.nix`
Signed-off-by: Jakub Sokołowski <jakub@status.im>
Changes:
- Drop `nix/mobile/android/maven-and-npm-deps/default.nix`
- Replace it with much simpler `nix/tools/patchNodeModules`
- Move Gradle patching tool to `nix/pkgs/patch-maven-srcs`
- Simplify it by using `gradle.deps` and patched node modules separately
- Change `TARGET` for `release-android` to `default`
- Move `mobile/reset-node_modules.sh` to `scripts/node_modules.sh`
- Move `nix/mobile/android/targets/release-android.nix` to `nix/mobile/android/release.nix`
Signed-off-by: Jakub Sokołowski <jakub@status.im>