Yarn install instructions

.circleci/config.yml

@@ -8,7 +8,8 @@ jobs:
       - checkout
       - restore_cache:
           key: tsickle-{{ checksum "yarn.lock" }}
-      - run: yarn install # don't bazel's yarn, since yarn install calls bazel, and bazel can't reenter
+      # Don't lock the bazel server, so that yarn can call bazel again
+      - run: bazel run @yarn//:yarn --script_path=yarn_install.sh && ./yarn_install.sh
       - run: bazel build ...
       - save_cache:
           key: tsickle-{{ checksum "yarn.lock" }}

.gitignore

@@ -9,3 +9,6 @@
 
 # Bazel working directories
 /bazel-*
+
+# See reference to this file in the README.md
+yarn_install.sh

README.md

@@ -31,10 +31,6 @@ builds that are seen outside of Google.
 We would like to make tsickle usable for everyone but right now if you'd like
 to try it you should expect to spend some time debugging and reporting bugs.
 
-## Installation
-
-- Execute `npm i` to install the dependencies.
-
 ## Usage
 
 ### Project Setup
@@ -102,6 +98,16 @@ Example:
 
 ## Development
 
+### One-time setup
+
+Run `bazel run @yarn//:yarn --script_path=yarn_install.sh && ./yarn_install.sh`
+to install the dependencies.
+
+> This avoids occupying the `bazel` server, so that `yarn` can call `bazel`
+> again.
+> Ideally we should just use `bazel-run.sh @yarn//:yarn`, see
+> https://stackoverflow.com/questions/47082298/how-can-users-get-bazel-run-sh
+
 ### Test commands
 
 - `ibazel test test:unit_test` executes the unit tests in watch mode (use `bazel test test:unit_test` for a