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

@@ -180,7 +180,7 @@ describe('.getOTPVersion(_) - Erlang', () => {

    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 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

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

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

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

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

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

Are you sure you want to change the base?

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

Uh oh!

Conversation

Uh oh!

Description

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

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

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