deniz.in

Markets

Weather

Loading weather

· via dev.to (home feed)

CocoaPods trunk stops accepting new podspecs on 2 December 2026; React Native migration shows the cost

On 2 December 2026 CocoaPods trunk stops accepting new podspecs. A hands-on React Native migration to Swift Package Manager finds six patches, one downgrade and a familiar set of failure modes.

CocoaPods trunk stops accepting new podspecs on 2 December 2026; React Native migration shows the cost

On 2 December 2026, CocoaPods trunk stops accepting new podspecs. That is the deadline framing a detailed migration write-up by Warren de Leon published on dev.to, which tests what actually happens when a real React Native app drops CocoaPods for Swift Package Manager.

The freeze itself is gentle. According to the post, existing builds keep working and pods that already resolve keep resolving. What ends is updates delivered through trunk, and only for the dependencies a project's Podfile.lock lists under its trunk source. The sharper pressure comes from libraries walking away from CocoaPods entirely: Firebase's migration page, as cited in the article, names Firebase 12 as the final major version published there, with new releases stopping in October 2026. An app that stays put risks pinning versions indefinitely.

React Native's escape hatch

In August, React Native 0.87 added an experimental Swift Package Manager path: a single command, run from the ios directory, that removes CocoaPods from an existing Xcode project (cd ios && npx react-native spm --deintegrate). CocoaPods remains React Native's default. The proposal behind the command, RFC #994, sets no date for dropping it and says existing apps "should not need to migrate immediately". An initial check in the tool reports which libraries lack a SwiftPM manifest and exits without touching the project, so it can be tried safely on a branch.

One app, twelve libraries

De Leon ran the migration on a finished app — a Pokédex — with twelve native libraries chosen to cover eight kinds of iOS dependency, from lottie-react-native and react-native-screens to react-native-svg, @notifee/react-native, react-native-permissions and react-native-mmkv running on Nitro Modules.

The outcome splits cleanly. Four libraries moved untouched. Five needed a patched manifest. One, the Material Design icons package, effectively stopped being a native module and left only its font in the app. The last two, react-native-mmkv and the Nitro Modules it runs on, could not make the trip: mmkv returned one major version older, with a patch of its own. The final cost is six patches to maintain and one downgrade, in exchange for a build that needs no Ruby on the machine.

What the patches replace

Each patch carries work that pod install used to do in Ruby: the permissions handler list, once rewritten into the podspec on disk, now written into the manifest; Lottie's package name set through swiftpmConfig.name instead of being computed at install time; the RCT_NEW_ARCH_ENABLED define added by hand, plus the Folly flags for mmkv; a header search path replacing a codegen header_dir; and a generated prefix header, force-included in every scaffold. One casualty is functional rather than structural: react-native-svg's precompiled Metal kernels were excluded, leaving two feComposite filter operators doing nothing.

Three further fixes come from SwiftPM's own rules rather than lost Ruby logic — the name of a URL-declared package, an umbrella-header layout CocoaPods never enforced, and a Swift access level one pod had hidden.

Six failure modes

Across the twelve builds the migration took, every failure fell into one of six categories: mismatched package or product names; headers or resources that SwiftPM handles differently; missing header paths or prefix headers CocoaPods used to supply; Swift and Objective-C code split into two targets; a package sync that runs after Xcode has already resolved; and a native module opting out of the build while its JavaScript still imports it. The manifest check surfaces the first problem; the builds expose the rest.

Timing and results

From first command to a feature-complete build took about 35 minutes and twelve builds, with testing and measurements filling the rest of the afternoon. A clean Release build came out roughly 50 seconds faster on one run each — a fresh-clone SwiftPM build timed at 159.2 seconds against 173.8 — though the author has not traced the cause. The app bundle and binary are smaller, and Ruby is gone from the build machine.

The caveats matter: everything ran on React Native 0.88.1 and Xcode 26.4.1, Release builds for the iPhone 17 Pro simulator, and the app was never run on a physical device. Nothing in the test set used vendored_frameworks, script phases, Expo modules or use_frameworks.

Why it matters

The December 2026 trunk freeze is an ecosystem deadline, not an app deadline — nothing breaks on the day. But as major libraries such as Firebase stop publishing pods, staying on CocoaPods becomes a slow version freeze, and the shared trunk stops being a place to publish anything new. This experiment suggests the React Native tooling is ready to attempt on a branch, and it is honest about the price: small manifest patches, one downgrade, a lost SVG filter, and untested territory for projects using vendored frameworks, script phases or Expo. Teams should start before their own dependencies stop updating, not after.

  • #react-native
  • #cocoapods
  • #swift-package-manager
  • #ios
  • #mobile-development