[SR-8213] Compiler may indicate "case will never be executed" even if it is unavoidable to execute the case in switch statement.

[YOCKOW]

Previous ID	SR-8213
Radar	None
Original Reporter	@YOCKOW
Type	Bug

Environment

• OS: macOS, Ubuntu 16.04

• Swift: 4.1.2, 4.2-DEVELOPMENT-SNAPSHOT-2018-07-08

Additional Detail from JIRA

Votes	0
Component/s	Compiler
Labels	Bug, DiagnosticsQoI, StarterBug
Assignee	Gabriel Igliozzi (JIRA)
Priority	Medium

md5: df5f44ab5b8a1d3c8ee6ed4a803d858c

is duplicated by:

• SR-11318 Incorrect "Case will never be executed" warning

Issue Description:

(As I commented on SR-8208,)
The compiler shows a wrong warning, when the following code is compiled:

enum E {
  case a // only one case
}

func f(_ e0:E, _ e1:E) {
  switch (e0, e1) {
  case (.a, _): print("e0 is .a") // -> No warning.
  }
  
  switch (e0, e1) {
  case (.a, _), (_, .a): print("e0 or e1 is .a") // -> warning: case will never be executed
  }
}

f(.a, .a)

Consideration

• The warning should be
  • case will be always executed
  • (_, .a) will never match
  • ... and so on

[belkadan]

@xedin, think this one's easy enough to be a StarterBug? You'd have to do some checking about whether a pattern was the only one in the case…

[xedin]

I think it makes sense to mark it as a StarterBug, it seems like that could be checked when cases are projected...

[vguerra]

hello @YOCKOW, @belkadan, @xedin 🙂

If none of you is working on this, I would like to tackle it.

[xedin]

Sure, go ahead, @vguerra!

[belkadan]

Resetting assignee for all Starter Bugs not touched since 2018.

[LucianoPAlmeida]

@belkadan @xedin There is someone working on this one? If not, I could give it a try

[swift-ci]

Comment by Gabriel Igliozzi (JIRA)

@xedin @belkadan If no-one is working on this bug I would like to give it a go as my first contribution

[xedin]

Maybe @LucianoPAlmeida is working on this, otherwise please feel free to take it over!

[LucianoPAlmeida]

I'm not working on this so go-ahead Gabriel Igliozzi (JIRA User) 🙂
But I think since this is linked as duplicated by an issue that is marked as resolved already, chances are that this is fixed as well.
So maybe the first step is to verify is the issue is still reproducible on the master branch

[swift-ci]

Comment by Gabriel Igliozzi (JIRA)

Hey @LucianoPAlmeida thanks for the response, I just did an pull on the master branch, compiled and ran that test code (with the swiftc command) and got

test.swift:11:3: warning: case will never be executed
  case (.a, _), (_, .a): print("e0 or e1 is .a") // -> warning: case will never be executed
  ^

Is this a valid way to test the master branch?, if so I believe the error is not yet resolved.

[LucianoPAlmeida]

Seems like the problem is still here, normally a quick and easy way to test if an issue is reproducible on master is with https://swift.godbolt.org there you can find the latest nightly snapshot from the master, but of course, you can download the lastest snapshot or build master locally and create a new file with the code as a test case and run it using lit 🙂

[jreference]

Should this issue be closed? In the current version of swift, no warnings are produced. Added a test case in #65819 to verify.

[AnthonyLatsis]

The warning itself was not a bug. It means that the second pattern will never be matched because a previous one always matches, which is true for case (.a, _), (_, .a):. The warning was silenced in #33336, but the issue is still relevant because we ought to diagnose the redundancy in the type checker (as in e.g. case .a, .a:) if we choose not to diagnose it further down the line.

[jreference]

@AnthonyLatsis Thank you for explaining!