Revert "IRGen: fix alignment for tail allocated arrays with elements which have a large alignment"

[rudkx]

Reverts #14652

Based on https://ci.swift.org/job/oss-swift-incremental-RA-linux-ubuntu-16_10/2703/, it seems like this might broken the stdlib/String.swift test on Linux.

What's curious is it passed PR testing, but two of my subsequent PR test attempts have failed in the same way. Perhaps this was a case of two commits landing at the same time, each passing individually but failing when combined?

[rudkx]

@swift-ci Please smoke test

[rudkx]

@eeckstein I'm doing a test run with this reverted to see if it fixes the CI failure (and PR testing failures).

[eeckstein]

I'll try to reproduce it locally

[eeckstein]

I couldn't reproduce it locally. Also it's very unlikely that my change would trigger this test failure

[eeckstein]

The test passed again in https://ci.swift.org/job/oss-swift-incremental-RA-linux-ubuntu-16_10/2705/
(but there is another failure afterwards)

[atrick]

Two failures have been blocking PR testing since this commit:

OS X Simulator:
[ FAIL ] NSStringAPIs/Substring.replacingCharacters(in:with:)

Linux:
[ FAIL ] StringTests.growth

The first CI failures was here:
https://ci.swift.org/job/oss-swift_tools-RA_stdlib-RD_test-simulator/671/

[atrick]

The String failure on linux seems to be sporadically failing on CI, even though my PR testing is consistently failing.

[rudkx]

Perhaps it this is related to test changes that @lancep made in 48018fb (around 5 builds before the first failure) which might have exposed a latent problem, or recent work by @milseman on Strings?

[atrick]

I think @lancep plans to revert and investigate the failures.

[rudkx]

Closing with the expectation that #14657 will resolve the failure.