WIP

apps/photos/.env.development

@@ -1,43 +1,36 @@
 # Sample configuration file
 #
 # All variables are commented out by default. Copy paste this into a new file
-# called `.env.local`, or create a new file with that name and add the
-# environment variables you need into it. That `.env.local` file will be
-# .gitignored, so you can freely customize how the app runs in your local setup.
+# called `.env.local` (or create a new file with that name) and add the
+# environment variables you want to apply during development. `.env.local` is
+# gitignored, so you can freely customize it for your local setup.
 #
-# - `.env.local` is picked up by next when NODE_ENV is development
+# `.env.local` is picked up by Next.js when NODE_ENV is 'development' (it is
+# 'production' by default, but gets set to 'development' when we run `next dev`)
 #
-# - `.env` is picked up always
-#
-# You don't necessarily need to use these files, these variables can be provided
-# as environment variables when running yarn dev too. e.g.
+# Alternatively, these variables can be provided as environment variables e.g.
 #
 #     NEXT_PUBLIC_ENTE_ENDPOINT=http://localhost:3000 yarn dev:photos
 #
-# Variables prefixed with NEXT_PUBLIC_ (in our case, all of them) are made
-# available when next runs our code in the browser (Behind the scenes, Next just
-# hardcodes occurrences of `process.env.NEXT_PUBLIC_FOO` with the value of the
-# `NEXT_PUBLIC_FOO` env var when the bundle is built). For more details, see
+# Variables prefixed with NEXT_PUBLIC_ are made available when Next.js runs our
+# code in the browser (Behind the scenes, Next.js just hardcodes occurrences of
+# `process.env.NEXT_PUBLIC_FOO` with the value of the `NEXT_PUBLIC_FOO` env var
+# when the bundle is built). See
 # https://nextjs.org/docs/pages/building-your-application/configuring/environment-variables
 #
-# By default, the app is configured to connect to the production instance etc.
-# This is usually a good default, for example someone might want to run the
-# client locally but still use their actual ente account.
+# By default, the app is configured to connect to the production APIs. This is
+# usually a good default, for example someone might want to run the client
+# locally but still use their actual ente account.
 #
-# Even though it connects to the production instances, when invoked with `yarn
-# dev:*`, next will behave as if NODE_ENV was set to 'development' (otherwise
-# this is assumed to be 'production'). There are some other cases too when we
-# assume we're in a dev environment (e.g. the NEXT_PUBLIC_APP_ENV env var below.
-# For the full list of rules that decide what counts as a dev build, see the
-# `isDevDeployment` function).
+# However, in other aspects it (by default) behaves like a development build
+# when executed using `yarn dev:foo`:
 #
-# We have some development related conveniences tied to the dev build:
 # 1. Logs go to the browser console instead of the log file
+#
 # 2. Sentry crash reporting etc is disabled
 #
-# The variables below thus serve as ways to customize which API instance we
-# connect to for various purposes. These variables are only honoured when we're
-# in a development environment.
+# See `isDevDeployment` for the exact rules for determining what counts as a
+# development build.
 
 # The ente API endpoint
 # NEXT_PUBLIC_ENTE_ENDPOINT=http://localhost:3000
@@ -62,39 +55,24 @@
 # Enhancement: Consider moving that into the app/ folder in this repository.
 # NEXT_PUBLIC_ENTE_FAMILY_PORTAL_ENDPOINT = http://localhost:3003
 
-# This in not useful when running locally. It is used to provide us a way to
-# mark certain deployments as "staging" by setting environment variables in the
-# CI job that deploys them on a remote server. See the `isDevDeployment`
-# function.
-#
-# By default, the photos web app gets deployed to "web.ente.io".
-# NEXT_PUBLIC_ENTE_WEB_ENDPOINT = http://localhost:3000
-
 # Set this to true to disable reporting of crashes to Sentry.
 #
 # Crash reporting is disabled if the user has opted out. This provides another
 # way to disable crash reporting, say for local test branches.
 # NEXT_PUBLIC_DISABLE_SENTRY=true
 
-# Set this to disable the upload of files via CF Workers
+# Set this to true to disable the upload of files via CF Workers
 #
 # CF workers provide us with a way of make the file uploads faster. The why and
 # how is explained here: https://ente.io/blog/tech/making-uploads-faster/
 #
 # By default, that's the route we take. This flag can be set to true to disable
-# that route, and instead directly upload to the S3-compatible URLs provided by
-# our API server.
+# that and instead directly upload to the S3-compatible URLs returned by the
+# ente API.
 #
-# Note the double negative.
+# Mind the double negative.
 # NEXT_PUBLIC_DISABLE_CF_UPLOAD_PROXY = true
 
-# This is an alternative to run as a development build.
-#
-# You likely don't need this if you're running on your machine, because when
-# invoked with `next dev` (as is the case for `yarn dev:photos` etc), next will
-# behave as if NODE_ENV was set to 'development'.
-# NEXT_PUBLIC_APP_ENV = development
-
 # The path of the JSON file which contains the expected results of our
 # integration tests. See `upload.test.ts` for more details.
 # NEXT_PUBLIC_EXPECTED_JSON_PATH = /path/to/dataset/expected.json

packages/shared/network/api.ts

@@ -72,8 +72,37 @@ export const getFamilyPortalURL = () => {
     return `https://family.ente.io`;
 };
 
-/*
-It's a dev deployment (and should use the environment override for endpoints ) in three cases:
+/**
+ * A build is considered as a development build if one of the following holds:
+ *
+ * 1. The NODE_ENV environment variable is set to 'development'. This
+ *    automatically happens when we run `yarn dev:foo`, but we can also
+ *    explictly set this to development before invoking the build. From the
+ *    Next.js docs:
+ *
+ *    > If the environment variable NODE_ENV is unassigned, Next.js
+ *    > automatically assigns development when running the `next dev` command,
+ *    > or production for all other commands.
+ *
+ * 2. Sometimes we're building for a remote deployment, but we want the deployed
+      site to behave as a development build. For example, when deploying the
+      main branch to `testing.ente.io`. In these cases, since the build was done
+      using `yarn export` (which in turn invokes `next build`), the NODE_ENV
+      will not get set to 'development'. To handle such cases, we introduce
+      another variable, NEXT_PUBLIC_ENTE_ENV, which has the same semantics as
+ *
+ *
+ * If the environment variable NODE_ENV is unassigned, Next.js automatically
+   assigns development when running the next dev command, or production for all
+   other commands.
+ *    next sets NODE_ENV to `production`.
+ *  When we run
+ *    `yarn dev:foo`, it invokes `next dev`, which sets NODE_ENV to
+ *    'development'. In all other cases (say, `next build`),
+ *
+
+It's a dev deployment (and should use the environment override for endpoints )
+in three cases:
 1. when the URL opened is that of the staging web app, or
 2. when the URL opened is that of the staging album app, or
 3. if the app is running locally (hence node_env is development)

scripts/deploy.sh

@@ -1,33 +1,46 @@
 #!/bin/sh
 
 # This script is run by the Cloudflare Pages integration when deploying the apps
-# in this repository. The app to build is decided based on the the value of the
-# CF_PAGES_BRANCH environment variable.
+# in this repository. The app to build and the environment variables to use is
+# decided based on the the value of the CF_PAGES_BRANCH environment variable.
 #
 # Ref: https://developers.cloudflare.com/pages/how-to/build-commands-branches/
-#
-# The CF Pages configuration is set to use `out/` as the build output directory,
-# so once we're done building we copy the app specific output to `out/`.
 
 set -o errexit
 set -o xtrace
 
+# The Cloudflare Pages build configuration is set to use `out/` as the build
+# output directory, so once we're done building we copy the app specific output
+# to `out/` (symlinking didn't work).
+
 rm -rf out
 
-if test "$CF_PAGES_BRANCH" = "auth-release"
-then
-    # By default, for preview deployments the NEXT_PUBLIC_APP_ENV is set to
-    # "test" in the CF environment variables. For production deployments of the
-    # auth app, reset this to "production".
-    #
-    # This is not needed for the default `yarn export:photos` case, because
-    # there the actual production deployment runs without NEXT_PUBLIC_APP_ENV
-    # being set to anything (and the other preview deployments have
-    # NEXT_PUBLIC_APP_ENV set to "test", as is correct).
-    export NEXT_PUBLIC_APP_ENV=production
+# Cloudflare Pages has two separate environments - Production and Preview.
+#
+# Each of these have their separate environment variables. However, we need to
+# deploy multiple production apps - so while for the "photos-release" branch
+# (which corresponds to Cloudflare's "Production" environment) can have separate
+# environment variables, the rest of the production deployments share the same
+# environment variables (those that are set for the Preview environment in CF).
+#
+# So we instead tune environment variables for specific deployments here.
+
+if test "$CF_PAGES_BRANCH" = "photos-release"; then
+    yarn export:photos
+    cp -R apps/photos/out .
+elif test "$CF_PAGES_BRANCH" = "auth-release"; then
     yarn export:auth
     cp -R apps/auth/out .
 else
+    # Apart from the named branches, everything else gets treated as a
+    # development deployment.
+    export NODE_ENV=development
+    # Also, we connect all of them to the dev APIs.
+    export NEXT_PUBLIC_APP_ENV=development
+    export NEXT_PUBLIC_ENTE_ENDPOINT=https://dev-api.ente.io
+    export NEXT_PUBLIC_ENTE_WEB_ENDPOINT=https://dev-web.ente.io
+    export NEXT_PUBLIC_ENTE_ALBUM_ENDPOINT=https://dev-albums.ente.io
+
     yarn export:photos
     cp -R apps/photos/out .
 fi