SR-10689: Fix bugs of DataProtocol's firstRange(of:in:)/lastRange(of:in:).

[YOCKOW]

DataProtocol's firstRange(of:in:) and lastRange(of:in:) return a wrong value or crash under some conditions because they handle indices incorrectly.
This PR fixes the issue.

Resolves SR-10689.

---

Note:
SR-10689 also says "lastRange(of:) returns firstRange(of:in:)", however, it has been already fixed by #2310.

[spevans]

@swift-ci test

[YOCKOW]

Rebased to resolve conflicts.

[spevans]

@swift-ci test

[YOCKOW]

The failures seem to be unrelated to this PR, don't they...?

[CodaFi]

@swift-ci test

[CodaFi]

@swift-ci test

[YOCKOW]

@CodaFi

Thank you for running tests.

The error on macOS seems to be unrelated.

[CodaFi]

@swift-ci test

[YOCKOW]

The failure on Linux seems to be caused by LLDB...

[YOCKOW]

I believe this is not stale... so rebased.
(Or we should wait for generalized "Collection Matchers"?)

[spevans]

@swift-ci test

[YOCKOW]

@spevans Thank you for rerunning tests.

---

The error seems to be unrelated.

/home/buildnode/jenkins/workspace/swift-corelibs-foundation-PR-Linux/swift/validation-test/StdlibUnittest/RaceTest.swift:116:11: error: CHECK: expected string not found in input
16:55:33 // CHECK: [ OK ] Race.closure
16:55:33 ^
16:55:33 :17:1: note: scanning from here
16:55:33 stderr>>> CRASHED: SIGSEGV
16:55:33 ^
16:55:33 :19:3: note: possible intended match here
16:55:33 [ FAIL ] Race.closure
16:55:33 ^
16:55:33
16:55:33 --

[spevans]

@swift-ci test linux

[YOCKOW]

@spevans Thank you again. At last tests have passed.

[YOCKOW]

Finally PR for SDK Overlay (swift#28639) has been merged.
I expect this PR is also ready to be merged.

[juozasvalancius]

I've just discovered that the previous implementation also had bugs in the returned length of the range.
So it's probably a good idea to add tests for that too.

Here is a case that returns shorter range than the given input in Xcode 11.6 (11E708):

import Foundation
let data = Data([1, 2, 3, 4, 5, 6])
let search = Data([2, 3, 9])
data.firstRange(of: search, in: 0 ..< 5) // returns 1..<2, should be nil

[YOCKOW]

@juozasvalancius
Thank you for the review.
The length is checked here https://github.com/apple/swift-corelibs-foundation/pull/2499/files#diff-593783312ec62e6e9299004a53468f83R865 and your case could be covered by https://github.com/apple/swift-corelibs-foundation/pull/2499/files#diff-593783312ec62e6e9299004a53468f83R892

[YOCKOW]

No conflicts even after the base branch was changed from master to main.

[YOCKOW]

Confirmation

• PR for SDK Overlay (swift#28639) was merged in Aug 2020. And now, the commit exists in this scf repository as a result of #2915. It is the same implementation with this PR.
• SR-10689 is not yet resolved on non-Darwin and will be resolved by this PR.
• All checks have passed, and this PR has no conflicts with main branch even now.
• (#2993 would resolve the same issue in the different way.)

---

@millenomi Could you review this PR (and/or #2993) at your convenience?

[CodaFi]

@swift-ci test

[YOCKOW]

@CodaFi
Thnak you for rerunning tests.
All tests have passed again.

[CodaFi]

I would like sign offs from some core maintainers here.

@millenomi @spevans

[millenomi]

Done.

[YOCKOW]

@millenomi Thank you for your approval.