ODBC uses correct null-tolerant rebSpellWideMaybe() #786

Workflow file for this run

.github/workflows/web-build.yml at 8516f0f

	#
	# File: %web-build.yml
	#
	#=============================================================================#
	#
	# GitHub Workflow for building the Wasm version of Ren-C. It builds all pull
	# requests to master, but only deploys commits that are actually merged in.
	#
	# Deployments will become "live" in the Web REPL demonstration directly from
	# this action...but only if they are "greenlit" after running a test in a
	# headless Firefox browser that is automated via Python and "Marionette".
	#
	#====# PLEASE READ THE README #===============================================#
	#
	# Whenever this file says "See README", that is referring to the notes in the
	# %.github/workflows/README.md file. If something appears in multiple GitHub
	# Workflow files, it's best to document it there instead of repeating it:
	#
	# https://github.com/metaeducation/ren-c/blob/master/.github/workflows/README.md
	#

	name: Emscripten Web


	# See README: When To Trigger Builds
	#
	on:
	push:
	branches: [
	master, # The web build is top priority, needs to always be checked!
	web # pushing to web won't build other workflows, use to debug
	]
	pull_request:
	branches: [
	master
	]
	workflow_dispatch: # Allows running this workflow manually from Actions tab


	# Standardize to use bash on all platforms.
	#
	# See README: Using The Strict Erroring Bash Shell
	#
	defaults:
	run:
	shell: bash


	# Each "Job" runs in its own VM, and a workflow run is made up of one or more
	# jobs that can run sequentially or in parallel.
	#
	# See README: Jobs
	#
	jobs:
	web-build: # Name of this workflow's only job

	# https://github.com/actions/virtual-environments#available-environments
	#
	runs-on: ubuntu-22.04


	# See README: Build Matrix
	#
	strategy:
	matrix:
	include:
	- os-id: 0.16.1 # "asyncify" Emscripten build (only variant ATM)
	config-file: emscripten.r


	# See README: Environment Variables
	#
	env:
	AWS_S3_BUCKET_NAME: metaeducation

	# See README: Minimize GitHub-Specific Syntax
	#
	OS_ID: ${{ matrix.os-id }}
	CONFIG_FILE: ${{ matrix.config-file }}


	# Steps are a sequence of tasks that will be executed within a single VM
	# as part of the job.
	#
	# See README: Steps
	#
	steps: # (no indentatation needed below; so indent the minimum!)


	#====# CHECKOUT STEPS #=====================================================#


	# https://github.com/actions/checkout
	#
	# See README: Checkout Action
	#
	- uses: actions/checkout@v4 # See README: Trusted Actions


	# The full commit is passed to make to build into the binary, and the
	# abbreviated commit is used to name the executable.
	#
	# See README: Portably Capturing Git Hashes
	#
	- name: Grab Git Hash and Short Hash Into Environment Variables
	run: \|
	git_commit="$(git show --format="%H" --no-patch)"
	git_commit_short="$(git show --format="%h" --no-patch)"
	echo "GIT_COMMIT=$git_commit" >> $GITHUB_ENV
	echo "GIT_COMMIT_SHORT=$git_commit_short" >> $GITHUB_ENV


	#====# TOOLCHAIN INSTALLATION STEPS #=======================================#

	# !!! Ideally this would use the same step that clients can use to build
	# the system with `make.sh`. Unfortunately, something about the GitHub
	# Ubuntus do not like the old bootstrap executable. Make sure the
	# ordinary path works, but for the moment patch over it just to get
	# to a point where the action works.
	#
	- name: Fetch R3 To Use For "Prep" Build Steps as $R3MAKE
	run: \|
	repo_dir=$(pwd)/
	source tools/bash/fetch-prebuilt.sh
	r3make=$(fetch_prebuilt)
	echo "R3MAKE is set to $r3make"
	echo "But that executable won't run on GitHub for some reason"
	# "$r3make" --do "print {TESTING 1 2 3}" # NOT WORKING, dunno why
	cd prebuilt
	wget http://hostilefork.com/media/shared/github/r3-linux-8994d23-patched
	chmod +x r3-linux-8994d23-patched
	r3make=$(pwd)/r3-linux-8994d23-patched
	echo "So now R3MAKE is $r3make"
	echo "R3MAKE=$r3make" >> $GITHUB_ENV # pass to next step


	- name: Stop the build early if the R3MAKE is no good
	run: \|
	"$R3MAKE" --do "print {R3MAKE is Working} quit"


	# This action will install the Emscripten SDK, which makes the `emcc`
	# compilation command available. This compiler takes in C files and will
	# emit wasm, for producing libr3.js
	#
	# https://github.com/mymindstorm/setup-emsdk
	#
	# See README: !!! IMPORTANT - Untrusted Actions, Use Audited Hash !!!
	#
	- uses: mymindstorm/setup-emsdk@d233ac12b0102f74ca199f5dad7a4e2c13a8a745
	with:
	version: 'latest-upstream'


	# Show a little bit of sanity check information
	#
	- name: Output System Information
	run: \|
	echo "Current directory is: $(pwd)"
	echo "EMCC version check:"
	emcc -v


	#====# BUILD STEPS #========================================================#


	# !!! optimization is currently hardcoded in the web build config files as
	# `s`. Review if `z` would be better (it cannot be passed in the options
	# here at time of writing, and would be ignored even if it could be due to
	# that ldflags config hardcoding).
	#
	# See README: {Braces} For %make.r String Parameters
	#
	# NOTE ON BUILDING WITH C++: We choose to build with C++ here instead of as
	# C, in order to use Emscripten's `-fwasm-exceptions` for native webassembly
	# support, in concert with REBOL_USES_TRY_CATCH=1.
	#
	# https://emscripten.org/docs/porting/exceptions.html#webassembly-exception-handling-proposal
	#
	# One reason to use this instead of Emscripten's setjmp()/longjmp() is that
	# emulation is not available on other Wasm runners like WasmEdge. Also it's
	# apparently pretty messy internally to do that emulation. It seems to run
	# fine in Chromium and Firefox with the webassembly native exceptions too,
	# so it seems a good cohice at time of writing (mid-2022).
	#
	# (The setjmp()/longjmp() code path is still the default in desktop builds
	# and gets plenty of exercise there, so another good reason to let the web
	# build test the variation.)

	- name: Generate Makefile for Emscripten-Based Build
	run: \|
	mkdir build
	cd build

	"$R3MAKE" ../make.r \
	config="../configs/$CONFIG_FILE" \
	target=makefile \
	standard=c++11 \
	os_id=$OS_ID \
	debug=none \
	git_commit="{$GIT_COMMIT}" \
	rigorous=no \
	static=no \
	extensions=""


	- name: Create Folders For Build Products (Compiler Won't Create Them)
	run: \|
	cd build
	make folders


	- name: Prep the Build By Making Various Auto-Generated .h and .c Files
	run: \|
	cd build
	make prep


	# https://github.com/actions/upload-artifact
	#
	- name: Optional Download of Prep Files Before They Can Cause Build Failure
	if: false # Change this to true to download a file
	uses: actions/upload-artifact@v2 # See README: Trusted Actions
	with:
	name: tmp-internals.h
	path: build/prep/include/tmp-internals.h


	- name: Compile and Link the C Sources to To Make .wasm and .js Files
	run: \|
	cd build
	make -j 2 # Linux GitHub Runners have 2 cores, use 2 jobs


	#====# UPLOAD STEPS #=======================================================#

	# We only want to build pull requests, we do not want to upload them to
	# the AWS server. Deployment should happen only once a commit has been
	# accepted and pushed to master. And then, it should only be greenlit
	# (to be the version the web console uses) if it passes the smoke test in
	# a headless browser.
	#
	# Unfortunately...there's no particularly great way to exit the steps
	# cleanly now if it's only a pull request. We can stop the steps, but
	# it would look like an error:
	#
	# https://github.com/actions/runner/issues/662
	#
	# So either we write one giant monolithic step, or every subsequent step
	# has to be qualified with an `if: github.ref == 'refs/heads/master'`.
	# Though the latter is not ideal, it's good enough for government work.
	#
	# Note: Steps will be stopped by default if any fail, but you can turn
	# a step back on with an `if: failure()` condition.


	# This action configures the AWS keys stored in GitHub's "Secrets" for
	# the repository so that `aws s3` allows us to do uploads, without needing
	# to publish any passwords publicly:
	#
	# https://github.com/aws-actions/configure-aws-credentials
	#
	# See README: Trusted Actions
	#
	- name: Configure AWS Credentials
	if: github.ref == 'refs/heads/master' # see notes on DEPLOY STEPS
	uses: aws-actions/configure-aws-credentials@v4
	with:
	aws-access-key-id: ${{ secrets.METAEDUCATION_AWS_ACCESS_KEY }}
	aws-secret-access-key: ${{ secrets.METAEDUCATION_AWS_SECRET_KEY }}
	aws-region: us-east-1


	# Here we upload the files to AWS, but we don't bump the special file
	# that is used by %load-r3.js to decide which library commit to use yet.
	#
	# The commit ID is included in the upload because the AWS bucket holds
	# many versions of the build products at once. To ask the web console to
	# use a specific commit ID, say:
	#
	# http://hostilefork.com/media/shared/replpad-js/?git_commit=<<shorthash>>
	#
	# The .js file contains the loading and setup code, which includes "cwrap"
	# functions that offer friendly JS function interfaces that take types
	# like JS strings, instead of being limited to the "integer heap address
	# parameters only" nature of Wasm functions.
	#
	# The .wasm file bundles both the compiled C code for the interpreter and
	# memory image (C constants, including compressed mezzanine code).
	#
	# There used to be more files, but this has streamlined over time...and
	# the pthread build is no longer supported so there's no `.worker.js`.
	# The only remaining potential build products would be debug files. (This
	# would change if extensions are built as their own "DLL"s.)
	#
	# Note: %load-r3.js is a weak link in the test-before-greenlight strategy.
	# There's not a way to select the singular loader, so any changes must be
	# deployed before we can test any build (even a non-"greenlit" one).
	#
	# !!! The repl could be adjusted to use an alternate load-r3.js, which
	# would mean not naming it statically in the HTML. To keep file variants
	# from accruing in the root directory, it could use some constant second
	# name (like %load-r3-preflight.js). Review.
	#
	- name: Upload Files To AWS (Don't Bump Version Used By load-r3.js Yet)
	if: github.ref == 'refs/heads/master' # see notes on UPLOAD STEPS
	run: \|
	ls -alF build # so you can see what was built

	local=build/libr3.js
	remote=s3://${AWS_S3_BUCKET_NAME}/travis-builds/${OS_ID}/libr3-${GIT_COMMIT_SHORT}.js
	aws s3 cp $local $remote

	# We have to set the MIME type on .wasm files, or the browser will not
	# load them as .wasm (it uses a fallback mechanism that is slower)
	#
	local=build/libr3.wasm
	remote=s3://${AWS_S3_BUCKET_NAME}/travis-builds/${OS_ID}/libr3-${GIT_COMMIT_SHORT}.wasm
	aws s3 cp $local $remote --content-type application/wasm

	# (not in subdir of 0.16.x because loader needs to pick between them)
	# See note above about how changes to this file undermines greenlight
	#
	local=extensions/javascript/load-r3.js
	remote=s3://${AWS_S3_BUCKET_NAME}/travis-builds/
	aws s3 cp $local $remote


	#====# TESTING STEPS #======================================================#

	# Check the deployment before "green-lighting" the %last-deploy.short-hash
	#
	# The ren-c-action is able to deploy a web browser and use the commit of
	# a non-greenlit hash. It does this via a local Firefox, which it talks
	# to through Python equipped with the "Marionette" protocol. (Ren-C can't
	# be used at time of writing, because it lacks websockets...which are
	# needed to remote-control Firefox).

	- name: LATEST-OF Smoke Test
	# if: github.ref == 'refs/heads/master' # see notes on UPLOAD STEPS
	if: false # !!! Broken at time of writing; need to bypass
	uses: metaeducation/ren-c-action@release
	with:
	web: true
	commit: ${{ env.GIT_COMMIT_SHORT }}
	timeout: 15
	screenshot: latest-of
	script: \|
	(url: latest-of)
	print ["Result was:" mold url]
	assert [url? url]


	- name: Watchlist Smoke Test
	# if: github.ref == 'refs/heads/master' # see notes on UPLOAD STEPS
	if: false # !!! Broken at time of writing; need to bypass
	uses: metaeducation/ren-c-action@release
	with:
	web: true
	commit: ${{ env.GIT_COMMIT_SHORT }}
	timeout: 15
	screenshot: watch
	script: \|
	x: 10
	watch x
	assert [10 = watch 1]


	- name: Redbol Smoke Test
	# if: github.ref == 'refs/heads/master' # see notes on UPLOAD STEPS
	if: false # !!! Broken at time of writing; need to bypass
	uses: metaeducation/ren-c-action@release
	with:
	web: true
	commit: ${{ env.GIT_COMMIT_SHORT }}
	timeout: 15
	screenshot: redbol
	script: \|
	redbol
	block: [b c]
	assert [[a b c d] = compose [a (block) d]]


	# Early on, @gchiu wrote some code to interoperate with a JS chess board.
	# Since we have that example, test that it works. We don't know what the
	# screen looks like, but we can check it loads and runs to completion with
	# the shortest possible gameplay.
	#
	- name: Test Chess GUI Example
	# if: github.ref == 'refs/heads/master' # see notes on UPLOAD STEPS
	if: false # !!! Broken at time of writing; need to bypass
	uses: metaeducation/ren-c-action@release
	with:
	web: true
	commit: ${{ env.GIT_COMMIT_SHORT }}
	timeout: 15
	screenshot: chess
	script: \|
	animate-game: do @chess
	assert [
	comment [https://en.wikipedia.org/wiki/Fool%27s_mate]
	<done> = animate-game [
	f2f3 e7e6
	g2g4 d8h4
	]
	]


	# If these scripts pass, we will consider the new wasm binary to be
	# suitable for greenlighting to be pulled by the ReplPad by default.
	#
	# Cypress runs all `.cy.js` files in the e2e ("End To End") directory.
	# That directory exists with a README.md saying where to get the .cy.js
	# files from, so we just go ahead and parse those URLs out here.
	#
	- name: Collect Cypress Tests from Repositories We Want to Keep Working
	uses: metaeducation/ren-c-action@release
	with:
	checked: true
	script: \|
	cd %tests/cypress/e2e/
	list: parse as text! read %README.md [
	collect some [
	keep ["*" space url! elide newline]
	\| thru newline
	]
	]
	for-each url list [
	filename: split-path url
	write filename (read url)
	]


	# NPM insists on there being a package-lock.json file, if not you get:
	#
	# "Error: Action failed. Missing package manager lockfile."
	#
	# While there may be good reasons for Node.js codebases to commit these
	# lock files to their repositories, this is not a Node.js project. Make
	# the lock file as part of the test process.
	#
	- name: Create package-lock.json file in Cypress test directory
	run: \|
	cd tests/cypress
	npm install


	# The Marionette test process has been replaced with Cypress, which
	# can run multiple browsers. See results here:
	#
	# https://dashboard.cypress.io/projects/wqxv1u/runs
	#
	# 1. Environment variables starting with CYPRESS_ are automatically
	# imported by Cypress. Alternately, there is a `with: env:` option
	# (that requires all environment variables to be on the same line)
	#
	# 2. The Cypress recording key allows us to upload videos and results
	# of the run to cypress.io, which is easier than having to locate
	# and retrieve "GitHub artifacts". It requires a Cypress account,
	# and there are 3 users per organization in the free tier.
	#
	# 3. We pass a GitHub token "to allow accurately detecting a build
	# vs. a re-run build". GitHub tokens are automatically generated.
	#
	# 4. By default, the Replpad will load the last greenlit hash from the
	# server. But we're trying to decide if a library version should be
	# greenlit, so we want the Cypress scripts to specifically ask the
	# loader to use an arbitrary commit. Passing it via an environment
	# variable that Cypress will proxy to the browser is easiest.
	#
	- name: Run Replpad using Chrome
	uses: cypress-io/github-action@v6
	with:
	working-directory: tests/cypress
	browser: chrome
	config-file: cypress.config.js
	record: true
	env: # [1]
	CYPRESS_RECORD_KEY: ${{ secrets.CYPRESS_RECORD_KEY }} # [2]
	GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }} # [3]
	CYPRESS_GIT_COMMIT_SHORT: ${{ env.GIT_COMMIT_SHORT }} # [4]
	# DEBUG: '@cypress/github-action' # verbose debug output
	# DEBUG: 'cypress:' # see runner logs (: is far too much output)


	#====# GREENLIGHT STEP (MARK THE UPLOADED BUILD AS CURRENT) #===============#

	# Each ${OS_ID} directory contains several builds for recent commits.
	# Since they are served statically from S3, we don't have a query to
	# serve the most recent one that successfully built (based on a date
	# or other property). So we write a file with a fixed name in that
	# directory to identify the last build...it can be obtained via a
	# CORS fetch() request.
	#
	# However, the upload might take a while...or fail part way through.
	# Hence, we make this the last file uploaded--so that when the browser
	# tries to fetch files for that ID, they'll all be there.
	#
	- name: Greenlight Build for load-r3.js To Use If Tests Passed
	if: github.ref == 'refs/heads/master' # see notes on UPLOAD STEPS
	run: \|
	cd build
	local=last-deploy.short-hash

	# -n option to echo means "no newline at end" (it's not a "text file"
	# so there is no standard enforcing that it have one...and it's
	# easier in the client to not have it)
	#
	echo -n "${GIT_COMMIT_SHORT}" > $local

	remote=s3://${AWS_S3_BUCKET_NAME}/travis-builds/${OS_ID}/$local
	aws s3 cp $local $remote # upload

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

ODBC uses correct null-tolerant rebSpellWideMaybe() #786

Workflow file

ODBC uses correct null-tolerant rebSpellWideMaybe() #786

Jobs

Run details

Workflow file for this run