Fix: more versioning around `nightly` and `maint`/`main` #359

paulo-ferraz-oliveira · 2025-07-03T18:15:10Z

Description

This'll address two issues:

obtaining known-branch versions (like main/maint)
obtaining non-specific version nightly

I have read and understood the contributing guidelines

I'll start by pushing tests that show the issue and then a fix that shouldn't affect any other tests.

Issue 1: https://github.com/erlef/setup-beam/actions/runs/16057710018/job/45316130218#step:5:54
Issue 2: https://github.com/erlef/setup-beam/actions/runs/16057710018/job/45316130218#step:5:67

Closes #356.
Closes #358.

⚠️ of note: stuff like 26 (if strict) was getting calculated as maint-26. It's no longer the case, but I think we can live with that (?)

1. stop trying to "coerce" versions on input 2. handle `OTP-` in one more place 3. handle `nightly`

paulo-ferraz-oliveira · 2025-07-03T18:48:49Z

.github/workflows/macos.yml

-          - otp-version: '27'
-            os: 'macos-15'
-            version-type: 'strict'
-          - otp-version: '26'
-            os: 'macos-15'
-            version-type: 'strict'


This shoulda been failing but wasn't. Guess it wasn't as strict as we intended...

Oh, it was getting calculated as maint-<otp-version>.

paulo-ferraz-oliveira · 2025-07-03T18:57:36Z

test/setup-beam.test.js

    it('is Ok for known linux version', async () => {
      before = simulateInput('version-type', 'strict')
-      spec = '26'
+      spec = 'maint-26'


This made no sense, but now we see it. Why would we want 26 to strict to be maint-26? These days this is opt-in, so I think some sanity was restored.

There is no strict 27 for macos-15, for example

paulo-ferraz-oliveira · 2025-07-03T19:02:46Z

test/setup-beam.test.js

+      spec = 'main-otp-28'
+      otpVersion = 'maint-28'
+      expected = 'main-otp-28'
+      await setupBeam.installOTP(otpVersion)
+      got = await setupBeam.getElixirVersion(spec, otpVersion)
+      assert.deepStrictEqual(got, expected)


Here's the test for #358.

paulo-ferraz-oliveira · 2025-07-03T19:02:54Z

test/setup-beam.test.js

+    spec = 'nightly'
+    otpVersion = '28'
+    expected = 'nightly'
+    got = await setupBeam.getGleamVersion(spec, otpVersion)
+    assert.deepStrictEqual(got, expected)


Here's the test for #356.

paulo-ferraz-oliveira · 2025-07-03T19:04:06Z

@lpil (the strict I told you about, before, is not required - and most likely for your use case wouldn't make sense), @SteffenDE, feel free to review.

wojtekmach · 2025-07-04T07:09:26Z

of note: stuff like 26 (if strict) was getting calculated as maint-26. It's no longer the case, but I think we can live with that (?)

definitely, I would considered that a bug that version-type: strict, otp-version: 26 pointed to maint-26, an unreleased version. It couldn't work on Windows since there are no windows builds for maint*. What does it point to now or does it error?

paulo-ferraz-oliveira · 2025-07-04T17:09:15Z

of note: stuff like 26 (if strict) was getting calculated as maint-26. It's no longer the case, but I think we can live with that (?)

definitely, I would considered that a bug that version-type: strict, otp-version: 26 pointed to maint-26, an unreleased version. It couldn't work on Windows since there are no windows builds for maint*. What does it point to now or does it error?

We're not testing it on Windows, I believe. I'll see about adding a test to verify the new behavior.

wojtekmach · 2025-07-04T17:20:42Z

Using this PR,

          - elixir-version: "1.18.4"
            otp-version: "26"
            version-type: 'strict'

resolves to:

Elixir:
Erlang/OTP 26 [erts-14.2.5.10] [source] [64-bit] [smp:4:4] [ds:4:4:10] [async-threads:1] [jit:ns]

Elixir 1.18.4 (compiled with Erlang/OTP 26)
OTP:
26.2.5.13

(https://github.com/wojtekmach/setup_beam_bug/actions/runs/16078422072/job/45378645101)

Is that expected?

paulo-ferraz-oliveira · 2025-07-04T17:23:26Z

I think the strict should resolve exactly to input (give or take a v and OTP- prefix), so if 26 is 26.2.5.13 it's wrong.

We'll test once per platform in similar conditions

wojtekmach · 2025-07-04T17:24:44Z

Agreed

paulo-ferraz-oliveira · 2025-07-04T18:11:39Z

Is it possible you forgot version-type? https://github.com/wojtekmach/setup_beam_bug/actions/runs/16078422072/job/45378645101#step:2:9 (I'm seeing loose there)

Edit: even if you didn't and I'm not seeing something, I'm pushing code so that we have a validation that strict is strict (alongside an exception for the case you mentioned - which unfortunately, at least at the moment, we can't be too strict on, until we wanna parse the error message).

wojtekmach · 2025-07-04T18:15:54Z

Yes that's what happened, my workflow file was wrong, apologies for any confusion. https://github.com/wojtekmach/setup_beam_bug/actions/runs/16078422072/workflow

paulo-ferraz-oliveira · 2025-07-04T18:16:52Z

No prob. As stated, if we have tests for specific situations we have less doubts.

paulo-ferraz-oliveira · 2025-07-10T15:19:34Z

dist/index.js

      .forEach((line) => {
-        const otpMatch = line
-          .match(/^([^ ]+)?( .+)/)[1]
-          .match(/^([^-]+-)?(.+)$/)


This one goes away.

paulo-ferraz-oliveira · 2025-07-10T15:20:45Z

I'm merging, for release, in the hopes:

this makes for better results
this includes enough tests around the changes
this marginally simplifies the implementation

paulo-ferraz-oliveira · 2025-07-10T15:34:18Z

This was released in v1.20.4.

paulo-ferraz-oliveira added 3 commits July 3, 2025 19:27

Test Elixir main-otp-28 / Erlang maint-28

2fbbd64

Test Gleam nightly / OTP 28

eab9765

Tweak versioning

03120e3

1. stop trying to "coerce" versions on input 2. handle `OTP-` in one more place 3. handle `nightly`

paulo-ferraz-oliveira force-pushed the fix/more-versioning branch from 74737e6 to 03120e3 Compare July 3, 2025 18:33

paulo-ferraz-oliveira commented Jul 3, 2025

View reviewed changes

paulo-ferraz-oliveira force-pushed the fix/more-versioning branch 2 times, most recently from 23af438 to 127dcbd Compare July 3, 2025 18:56

paulo-ferraz-oliveira commented Jul 3, 2025

View reviewed changes

Remove tests that should be failing previously

3ecdae9

There is no strict 27 for macos-15, for example

paulo-ferraz-oliveira force-pushed the fix/more-versioning branch from 127dcbd to 3ecdae9 Compare July 3, 2025 18:59

paulo-ferraz-oliveira commented Jul 3, 2025

View reviewed changes

paulo-ferraz-oliveira requested a review from wojtekmach July 3, 2025 19:03

paulo-ferraz-oliveira marked this pull request as ready for review July 3, 2025 19:04

paulo-ferraz-oliveira requested a review from starbelly July 3, 2025 19:04

Remove similar tests

c6e0209

We'll test once per platform in similar conditions

paulo-ferraz-oliveira force-pushed the fix/more-versioning branch 5 times, most recently from 9b94d01 to b44ebc6 Compare July 4, 2025 17:49

await for assert.rejects (not noticeable if alone inside it)

f5509d6

paulo-ferraz-oliveira force-pushed the fix/more-versioning branch from b44ebc6 to f984a85 Compare July 4, 2025 18:10

Test maint- for strict and loose

0b1e9cb

paulo-ferraz-oliveira force-pushed the fix/more-versioning branch from f984a85 to 0b1e9cb Compare July 4, 2025 18:16

paulo-ferraz-oliveira commented Jul 10, 2025

View reviewed changes

paulo-ferraz-oliveira merged commit fceaea9 into main Jul 10, 2025
74 checks passed

paulo-ferraz-oliveira deleted the fix/more-versioning branch July 10, 2025 15:20

Fix: more versioning around nightly and maint/main #359

Fix: more versioning around nightly and maint/main #359

Uh oh!

Conversation

paulo-ferraz-oliveira commented Jul 3, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Uh oh!

paulo-ferraz-oliveira Jul 3, 2025

Choose a reason for hiding this comment

Uh oh!

paulo-ferraz-oliveira Jul 3, 2025

Choose a reason for hiding this comment

Uh oh!

paulo-ferraz-oliveira Jul 3, 2025

Choose a reason for hiding this comment

Uh oh!

paulo-ferraz-oliveira Jul 3, 2025

Choose a reason for hiding this comment

Uh oh!

paulo-ferraz-oliveira Jul 3, 2025

Choose a reason for hiding this comment

Uh oh!

paulo-ferraz-oliveira commented Jul 3, 2025

Uh oh!

wojtekmach commented Jul 4, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

paulo-ferraz-oliveira commented Jul 4, 2025

Uh oh!

wojtekmach commented Jul 4, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

paulo-ferraz-oliveira commented Jul 4, 2025

Uh oh!

wojtekmach commented Jul 4, 2025

Uh oh!

paulo-ferraz-oliveira commented Jul 4, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

wojtekmach commented Jul 4, 2025

Uh oh!

paulo-ferraz-oliveira commented Jul 4, 2025

Uh oh!

paulo-ferraz-oliveira Jul 10, 2025

Choose a reason for hiding this comment

Uh oh!

paulo-ferraz-oliveira commented Jul 10, 2025

Uh oh!

Uh oh!

paulo-ferraz-oliveira commented Jul 10, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Fix: more versioning around `nightly` and `maint`/`main` #359

Fix: more versioning around `nightly` and `maint`/`main` #359

paulo-ferraz-oliveira commented Jul 3, 2025 •

edited

Loading

wojtekmach commented Jul 4, 2025 •

edited

Loading

wojtekmach commented Jul 4, 2025 •

edited

Loading

paulo-ferraz-oliveira commented Jul 4, 2025 •

edited

Loading