Fix Issue 19549 - -check=in=off doesn't work

[wilzbach]

Copy over the existing tests he said.

And again he said please copy over the existing tests.

And he said again there are still no tests.

CC @andralex @thewilsonator PLEASE PLEASE STOP MERGING STUFF WITHOUT TESTS.

With the fixes, the first test file yields a segfault of DMD, because this assertion no longer holds:

dmd/src/dmd/s2ir.d

Line 426 in a9eb1b7

assert(s.label.statement);

but I'm not in the mood to look into this.

[dlang-bot]

Thanks for your pull request, @wilzbach!

Bugzilla references

Auto-close	Bugzilla	Severity	Description
✓	19549	critical	-check=in=off doesn't work

Testing this PR locally

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

dub fetch digger
dub run digger -- build "stable + dmd#9195"

[thewilsonator]

Yeah the original code look very broken because if (!global.params.useIn) is true any any explicit value of the switch, i.e. in=off has the same effect as in=on! This patch doesn't take into account @safe functions however

if (global.params.useOut == CHECKENABLE.off)

needs to be something like

if (global.params.useOut == CHECKENABLE.off ||
   (global.params.useOut == CHECKENABLE.safeOnly && f.trust != TRUST.safe))

[WalterBright]

Once it passes the test suite.

[wilzbach]

but I'm not in the mood to look into this.

Okay, I found the missing bit, but PLEASE PLEASE PLEASE let's NOT merge new features without tests again.

Yeah the original code look very broken

Yup any kind of test would have spotted that (or you know also no implicit bool conversion from an enum in a condition).

This patch doesn't take into account @safe functions however

CHECKENABLE.safeOnly is only used for useArrayBounds (a bit confusing to the same type for both though).

[thewilsonator]

Okay, I found the missing bit, but PLEASE PLEASE PLEASE let's NOT merge new features without tests again.

Duly noted.