Also remove the engine experiment, because that never really went
anywhere for canvas and was just one more "different" thing
[fsc-max-nodes=15]
[fsc-timeout=45]
Change-Id: Ib990deb94bf5e25d587d554f64c535c9b96cce26
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/314469
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
Reviewed-by: Isaac Moore <isaac.moore@instructure.com>
Reviewed-by: Aaron Ogata <aogata@instructure.com>
Build-Review: Aaron Ogata <aogata@instructure.com>
Migration-Review: Aaron Ogata <aogata@instructure.com>
QA-Review: Jacob Burroughs <jburroughs@instructure.com>
Product-Review: Jacob Burroughs <jburroughs@instructure.com>
refs DE-1487
the deploy build excludes the spec directory and tar fails with exit code 2 on this step. docker/dockerfile:1.5.2 now fails when tar fails with exit code 2, unlike docker/dockerfile:1.0-experimental.
Change-Id: I13ae5b306c6e94dbc59e1b5dac90d2816ca2054f
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/313345
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
Reviewed-by: Andrea Cirulli <andrea.cirulli@instructure.com>
QA-Review: Aaron Ogata <aogata@instructure.com>
Product-Review: Aaron Ogata <aogata@instructure.com>
Build-Review: Aaron Ogata <aogata@instructure.com>
refs AE-1492
Test Plan:
1. Ensure that all appropriate gem versions are installed for CANVAS_RAILS=6.1 with plugins enabled
2. Ensure that all appropriate gem versions are installed for CANVAS_RAILS=7.0 with plugins enabled
3. Ensure that removing Gemfile.rails*.lock and running “bundle install” produces both 6.1 & 7.0 lockfiles with no overwrites to the partial files
4. Ensure that all lockfiles update when using “bundle update” in both versions and there are changes to commit
5. Ensure that removing plugins and running “bundle install” produces no changes to commit
6. Ensure that “bundle exec rails c” with CANVAS_RAILS=6.1 runs the right version
7. Ensure that “bundle exec rails c” with CANVAS_RAILS=7.0 runs the right version
8. Ensure that the Gemfile.*.lock.partial does not contain any secrets or private plugin names
NOTE: “vcr” gem is intentionally pinned to an old version (6.0.0) for testing
Change-Id: Id8e449466ec7839b3b71f2430256c0e8d0967775
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/309093
Reviewed-by: Isaac Moore <isaac.moore@instructure.com>
Reviewed-by: Alex Slaughter <aslaughter@instructure.com>
QA-Review: Isaac Moore <isaac.moore@instructure.com>
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
Build-Review: Alex Slaughter <aslaughter@instructure.com>
Product-Review: Aaron Ogata <aogata@instructure.com>
refs DE-1379
The Jenkins checkoutRepo() command produces different file permissions than `git archive`, which caused all plugin builds to not use the same cache that the rest of the builds were using.
Change-Id: Icda3110eb3f2366ad846dd43a4e79551065a3a75
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/302239
Reviewed-by: Andrea Cirulli <andrea.cirulli@instructure.com>
QA-Review: Aaron Ogata <aogata@instructure.com>
Product-Review: Aaron Ogata <aogata@instructure.com>
Build-Review: Aaron Ogata <aogata@instructure.com>
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
refs DE-1379
Test Plan
1. JS Impact Analysis works
2. Tarball hashes are reproducible crossbuild
[build-registry-path=jenkins/canvas-lms/de-1379-3]
[change-merged]
Change-Id: I69631a80dd2cbdc03467b1ae690c0eeeaeda555c
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/301748
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
Reviewed-by: Bobby Buten <bobby.buten@instructure.com>
Build-Review: Bobby Buten <bobby.buten@instructure.com>
QA-Review: Aaron Ogata <aogata@instructure.com>
Product-Review: Aaron Ogata <aogata@instructure.com>
Refs FOO-2929
flag=none
The replaces some of the functionality performed in the
ui/index.js entrypoint with the new "engine" which will
eventually hold all the startup helpers for standalone
entrypoints (that don't boot via ui/index.js). For
starters, we've done the loading of the translation file,
the backfill of ENV.LOCALE from the ENV.LOCALES list,
and the (possible) loading of locale polyfills for the
Intl subsystems that are in use.
Test plan:
* Canvas pages come up normally
* Try a locale that requires polyfilling
(recommend Welsh (Cymraeg) in Chrome browser)
* That should still work, you should see console
messages affirming that the locale was polyfilled
for the three Intl subsystems that we use
Change-Id: Ie9b386d5006a03372583240afca06190a4f23ee2
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/294705
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
Reviewed-by: Jacob Burroughs <jburroughs@instructure.com>
QA-Review: Charley Kline <ckline@instructure.com>
Product-Review: Charley Kline <ckline@instructure.com>
refs FOO-2801
flag = none
[change-merged][build-registry-path=jenkins/canvas-lms/foo-2801]
gems/canvas_i18nliner is now a package @instructure/i18nliner-canvas and
lives in the same repo on github along with the 3 other i18nliner
libraries.. this was done to make it easier for maintainers to deal with
this code, since changing one part may break the other due to how
they're architected
the source on github: https://github.com/instructure/i18nliner-js
~ test plan ~
build is still OK, this only affects the generation of files, and those
i manually verified to be identical before and after
Change-Id: I78afa8a808f1699c10aced8466cfade066848bc9
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/294209
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
Reviewed-by: Charley Kline <ckline@instructure.com>
QA-Review: Charley Kline <ckline@instructure.com>
Product-Review: Charley Kline <ckline@instructure.com>
Change-Id: I61d5ba9f6ef5e0837da2ec22b695b9a56b88acb5
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/293104
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
Reviewed-by: August Thornton <august@instructure.com>
Reviewed-by: Alex Slaughter <aslaughter@instructure.com>
QA-Review: Jacob Burroughs <jburroughs@instructure.com>
Product-Review: Jacob Burroughs <jburroughs@instructure.com>
refs DE-523
Previously, the current version would be represented by Gemfile.lock and the next version by Gemfile.lock.next. Replace both with a common format of Gemfile.rails<version>.lock to be able to iterate over the files in a cleaner way.
Change-Id: I35aef3a14e726eb35db8aebc808af4a925552c01
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/288563
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
Reviewed-by: Jacob Burroughs <jburroughs@instructure.com>
QA-Review: Aaron Ogata <aogata@instructure.com>
Product-Review: Aaron Ogata <aogata@instructure.com>
refs FOO-2697
flag = none
those two packages were relying on canvas's babel.config.js but no more
test plan: the build phase of the CI is enough as there were no logical
changes
Change-Id: I7814796f36f1e30b4dd494a12f768f9dbafb4f1f
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/286759
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
Reviewed-by: August Thornton <august@instructure.com>
Product-Review: August Thornton <august@instructure.com>
QA-Review: Ahmad Amireh <ahmad@instructure.com>
minor QOL mainly for me; put webpack-specific configuration in its own
folder, make it portable, and make room for esbuild
the __webpack_public_path__ initializer no longer sources the value from
the build module, instead it uses a pre-defined global that webpack
injects
~ test plan ~
build is OK
Change-Id: I4ba4a3c0cb9175f96096f2b78022e152c04fc75d
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/276183
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
Reviewed-by: Charley Kline <ckline@instructure.com>
QA-Review: Charley Kline <ckline@instructure.com>
Product-Review: Charley Kline <ckline@instructure.com>
Psych has safe_load now, and it's fairly trivial to convert our existing
overrides to use that instead
Change-Id: I2648df8d4574e15fc9072a25882e318d902765c3
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/271939
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
Reviewed-by: Jacob Burroughs <jburroughs@instructure.com>
QA-Review: Cody Cutrer <cody@instructure.com>
Product-Review: Cody Cutrer <cody@instructure.com>
fixes FOO-1824
flag = none
[pin-commit-analytics=7f2f049830a11e43f4a5d0031e3c4ecc3145ccdb]
[pin-commit-instructure_misc_plugin=9b505159130c0724ce6bcbb75d07d40ff16a1961]
read directly from gems/plugins/* instead of symlinks
- dropped dead brandable_css config property "all_sass_files"
- webpack test runner now generates a file that runs all plugin spec
files (from spec_canvas/coffeescripts)
test plan: you can remove your symlinks and still survive!
Change-Id: I0206c2d827aa9f59b0374b21f0863443dff3be0f
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/262346
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
Reviewed-by: Michael Ziwisky <mziwisky@instructure.com>
Reviewed-by: Simon Williams <simon@instructure.com>
QA-Review: Simon Williams <simon@instructure.com>
Product-Review: Simon Williams <simon@instructure.com>
refs FOO-1408
flag = none
This reverts commit c904d83673.
Original commit 9af1badd2f.
I've also made an adjustment where you may only include a directory and
not a specific file so that we don't make it possible to configure
outside of an ".i18nrc" file (a measure to eliminate surprises that's
all)
before:
{ "include": [ "path/to/dir/.i18nrc" ] }
now:
{ "include": [ "path/to/dir" ] }
:: test plan
run "rake canvas:compile_assets" and visit any page with JavaScript
text, that text should display correctly
Change-Id: I320ec4056270647835035f45501d310b18d9cb25
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/258178
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
Reviewed-by: Michael Ziwisky <mziwisky@instructure.com>
QA-Review: Michael Ziwisky <mziwisky@instructure.com>
Product-Review: Michael Ziwisky <mziwisky@instructure.com>
refs FOO-1644
[fsc-propagate=false]
- create engines directory
- put one empty-ish engine in there
- add general engines dependency to
gemfile construction
- update build pipeline to run tests all engines
just like gems
- add engines directory to jenkins-cache build
- get rspec running successfully with a dummy app
inside the audit engine
- stop failing with no message if environment
unloadable in css config
- generally set pattern in build artifacts and in
code organization for pulling out an engine
Change-Id: I8650f5e0aa7d72556d2f02156886d2b503a111fb
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/258988
Reviewed-by: Jacob Burroughs <jburroughs@instructure.com>
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
QA-Review: Ethan Vizitei <evizitei@instructure.com>
Product-Review: Ethan Vizitei <evizitei@instructure.com>
refs FOO-1627
flag=none
make Dockerfile cache use
gem instead
TEST PLAN:
1) specs run successfully
2) config in operational environments loads successfully
3) reloading settings still causes config files to get
re-parsed
Change-Id: I69bdfc2b79065600336f83c7ffb87dd8e647fa9c
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/259232
Reviewed-by: Simon Williams <simon@instructure.com>
QA-Review: Simon Williams <simon@instructure.com>
Product-Review: Simon Williams <simon@instructure.com>
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
you can switch back to rails 5.2 one of these ways:
* ENV CANVAS_RAILS6_0=0
* consul key rails6.0 with value `false`
* touch config/RAILS5_2
Change-Id: I811884b2250b865220a1a9d84c7254f42d79e18f
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/257587
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
Reviewed-by: Rob Orton <rob@instructure.com>
Reviewed-by: Michael Ziwisky <mziwisky@instructure.com>
QA-Review: Michael Ziwisky <mziwisky@instructure.com>
Product-Review: Michael Ziwisky <mziwisky@instructure.com>
This reverts commit 9af1badd2f.
Change-Id: I159a478d1f3baa062148e3fd5ff4ee914176579c
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/256982
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
Reviewed-by: Simon Williams <simon@instructure.com>
QA-Review: Simon Williams <simon@instructure.com>
Product-Review: Simon Williams <simon@instructure.com>
refs FOO-1331
refs LS-1657
flag = none
test plan: canvas-rce mocha tests work after upgrading to inst-ui 7
Change-Id: Ibe8d94fdd5a994fda2c7b9703dccfb77a62415ca
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/256766
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
Reviewed-by: Ed Schiebel <eschiebel@instructure.com>
QA-Review: Ed Schiebel <eschiebel@instructure.com>
Product-Review: Simon Williams <simon@instructure.com>
fixes FOO-1409
flag = none
no more client_apps, canvas_quizzes now lives as part of canvas-lms
proper inside app/jsx/, which makes the build leaner and leaves us with
one less thing to reason about
logical changes:
- converted from AMD to ES modules
- upgraded to recent react + react-router
- dropped RSVP in favor of native Promises
- used CanvasModal instead of home-grown Dialog
- removed dead code; notifications in particular were fishy as there had
no dependents at all and did not even show up in the graph
- ported tests to Jest, added more unit ones and two integration ones
- removed "config.onError" and now throws errors where appropriate
- disabled console statements in non-dev
:: test plan ::
- create a (old-school) quiz containing all types of questions
- as 3 distinct students, take the quiz and try to randomize your
answers
at this point it's helpful to have a reference to compare the screens; I
replicated the quiz on my production sandbox for this
- go to /courses/:id/quizzes/:id/submissions/:id/log
- verify it looks OK
- click on a specific question in the stream and verify the question
inspector widget works OK
- go back to stream and push "View table"
- verify the table and its controls are OK
- go to /courses/:id/quizzes/:id/statistics
- verify it looks OK
- click on ? in the discrimination index chart and verify it displays
a dialog with help content
- click on "X respondents" in one of the charts and verify it displays
a dialog with the respondent names
- verify the interactive charts do interact as expected (no logic
changed here so just a quick glance)
- link to "View in SpeedGrader" for essay-like questions works
Change-Id: I79af5ff4f1479503b5e2528b613255dde5bc45d3
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/256118
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
Reviewed-by: Simon Williams <simon@instructure.com>
QA-Review: Simon Williams <simon@instructure.com>
Product-Review: Simon Williams <simon@instructure.com>
fixes FOO-1408
this allows us to control which files i18nliner processes completely
through config and without having to modify hardcoded paths in source
file, which is something we need for FOO-1265
.i18nrc files can include other directories through the "include"
directive:
// file: canvas-lms/.i18nrc
{ "include": [ "public/javascripts/.i18nrc" ] }
// file: public/javascripts/.i18nrc
{ "files": [...] }
:: test plan
aside of Jenkins exercising the i18n tasks, I ran a diff by hand over
the set of files that i18nliner processses before and after the patch,
with the new code processing a few more files: some handlebars in the
analytics plugin and the 3 client_apps/canvas_quizzes source files
if you really want to, you can do the same or find another way to verify
the output
on master, edit canvas_i18nliner/js/main.js somewhere before the
exports:
Check.prototype.checkWrapper = f => console.warn(f)
run it:
./gems/canvas_i18nliner/bin/i18nliner check 2>
tmp/i18nliner-upstream-files.txt
cat tmp/i18nliner-upstream-files.txt | sort >
tmp/i18nliner-upstream.txt
now do similar on our branch (although we need to massage teh output
because the paths are absolute:)
./gems/canvas_i18nliner/bin/i18nliner check 2>
tmp/i18nliner-patched-files.txt
cat tmp/i18nliner-patched-files.txt | sort >
tmp/i18nliner-patched.txt
sed -i "s{$PWD/{{" tmp/i18nliner-patched.txt
now look for differences:
git diff --no-index \
tmp/i18nliner-upstream.txt \
tmp/i18nliner-patched.txt
Change-Id: Ic73cbc7261ab597deb567fc5d0af1e3014875da1
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/255952
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
Reviewed-by: Simon Williams <simon@instructure.com>
QA-Review: Simon Williams <simon@instructure.com>
Product-Review: Simon Williams <simon@instructure.com>
fixes FOO-1116
flag = none
this upgrade was problematic mostly because of Node's support for ES
modules (ESM) that is still fairly recent and introduces some complexity
around packaging. This is also the reason for the considerable hack
related to @instructure packages (see inline documentation) that can be
removed once we fix things upstream
| test plan |
| --------- |
- install node 12 if needed (e.g. not on docker)
- purge your node modules:
rm -rf {.,client_apps/*,gems/*,gems/plugins/*,packages/*}/node_modules
- make sure you can still build locally (native/docker)
bundle exec rake canvas:compile_assets
Change-Id: If9605ac428f0ff228f8852f0becb9fbac750f1f2
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/253166
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
Reviewed-by: Simon Williams <simon@instructure.com>
Reviewed-by: Charley Kline <ckline@instructure.com>
QA-Review: Simon Williams <simon@instructure.com>
QA-Review: Charley Kline <ckline@instructure.com>
Product-Review: Charley Kline <ckline@instructure.com>
refs DE-376
For the ruby-runner change, we will need to have the yarn packages dependencies pre-computed in a separate docker image so we can collect the MD5SUM of those changes. Also, the final ruby-runner image will not contain the packages/ directory, so we have to move it to the webpack-builder image for the JS jobs.
Change-Id: I37698e06090dcbee815651120f92bf39538e886f
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/253584
Reviewed-by: James Butters <jbutters@instructure.com>
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
QA-Review: Aaron Ogata <aogata@instructure.com>
Product-Review: Aaron Ogata <aogata@instructure.com>
Buildkit seems to have a race condition where the cache is not used, or incompletely used, mostly on a new instance, but sometimes on an instance that has already used the cache successfully. We created a partial reproduction (g/250409) where with buildkit, layers above the modified layer were rebuilt, and with this change, all cached layers are correctly used.
[change-merged]
[build-registry-path=jenkins/canvas-lms/final-no-buildkit]
Change-Id: If10c186583cc21421d861bc2d1d43c7d2b0fc2f9
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/250400
Reviewed-by: James Butters <jbutters@instructure.com>
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
QA-Review: Aaron Ogata <aogata@instructure.com>
Product-Review: Aaron Ogata <aogata@instructure.com>
Aaron Ogata