fix Issue 21772 - [REG2.069] Consecutive different-signed double.nans…

[WalterBright]

… in an array literal take the sign of the previous nan (same for float and real)

[dlang-bot]

Thanks for your pull request, @WalterBright!

Bugzilla references

Auto-close	Bugzilla	Severity	Description
✓	21772	regression	[REG2.069] Consecutive different-signed double.nans in an array literal take the sign of the previous nan (same for float and real)

⚠️⚠️⚠️ Warnings ⚠️⚠️⚠️

• Regression or critical bug fixes should always target the stable branch. Learn more about rebasing to stable or the D release process.

To target stable perform these two steps:

1. Rebase your branch to upstream/stable:

git rebase --onto upstream/stable upstream/master

2. Change the base branch of your PR to stable

Testing this PR locally

If you don't have a local development environment setup, you can use Digger to test this PR:

dub run digger -- build "master + dmd#14727"

[dkorpel]

Bot says regressions should target stable

[ibuclaw]

Testing all versions of gdc from 9..12, I don't get the same bug as dmd here, CTFE values look fine too.

$ gdc r.d -o gdc-r
[-QNaN, +QNaN, +QNaN, 1.0e+0, +QNaN, -QNaN]
$ ./gdc-r
[-nan, nan, nan, 1, nan, -nan]

vs. dmd

$ dmd r.d -o dmd-r
[-nan, nan, nan, 1.0, nan, -nan]
$ ./dmd-r
[-nan, -nan, -nan, 1, nan, nan]
core.exception.AssertError@r.d(11): Assertion failure

So I'm not entirely sure if the CTFE changes are required here for fixing this bug.

Also bearing in mind that CTFE nan !is -nan was fixed in #13780.

[dkorpel]

As far as I understand, the issue is not actually a regression. It was always wrong, writlen just didn't print -nan before 2.069