Fix Issue 18828 - [-betterC] helpless error in object.d

[JinShil]

I don't really like this fix, but I'm not sure how else to do it, and maybe this PR will elicit some ideas from others.

[dlang-bot]

Thanks for your pull request, @JinShil!

Bugzilla references

Auto-close	Bugzilla	Severity	Description
✓	18828	blocker	[-betterC] helpless error in object.d

⚠️⚠️⚠️ 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 fetch digger
dub run digger -- build "master + druntime#2178"

[JinShil]

I'm not sure how to test -betterC in the runtime, so, instead, I added a test at dlang/dmd#8235

[jacob-carlborg]

Is this testable?

[JinShil]

Is this testable?

Yes, at dlang/dmd#8235. I didn't know how else to do it.

[jacob-carlborg]

Yes, at dlang/dmd#8235. I didn't know how else to do it.

Is this related to how the tests are setup or run in druntime? Because I don't see anything special in that test.

[JinShil]

Notice this at the top of the test file:

 /* REQUIRED_ARGS: -betterC

This means the test will be compiled in -betterC mode. In -betterC you cannot use TypeInfo, so using typeid will cause the compiler to emit an error. The current implementation of destroy (prior to this PR) uses typeid, which is invalid in -betterC.

You can see a demonstration of the test currently failing at https://run.dlang.io/is/vPcZNA. Once this is pulled the DMD test should pass.

[jacob-carlborg]

Not sure about how much work it would be, but what about adding a new directory here: https://github.com/dlang/druntime/tree/master/test ?

[JinShil]

I added the directory and test to the best of my ability, but I don't even understand what most of the syntax in the Makefile is for and I'm not even sure if the test is running.

EDIT: See comment below.

[jacob-carlborg]

I added the directory and test to the best of my ability, but I don't even understand what most of the syntax in the Makefile is for and I'm not even sure if the test is running.

Add assert(false); to verify that the tests are running and failing.

[JinShil]

It appears to be running now. I had to add the test to the root posix.mak file and output appeared (e.g. https://auto-tester.puremagic.com/show-run.ghtml?projectid=1&runid=3176138&dataid=22521943&isPull=true search for "betterc").

It appears none of the tests in the test folder are run for Windows except uuid. I think that's ok for this test as there's nothing platform specific about it.

[RazvanN7]

Can we close : dlang/dmd#8235 now ?

[schveiguy]

Haha, I just went into bugzilla to make a new issue about the inefficiency of destroy, to find that the code I swore was using typeid is no longer there. And lo and behold, it's changed, but for another reason.

I'm still going to file an issue, because I feel like smaller structs can be done much more efficiently than casting to ubyte arrays. But this change helps significantly.

See: https://forum.dlang.org/post/pe3pcp$13t4$1@digitalmars.com

[schveiguy]

I'm still going to file an issue

See https://issues.dlang.org/show_bug.cgi?id=18899