[dmd-cxx] fix Issue 19912: No implicit import of object module when an object declaration exists.

[ibuclaw]

No description provided.

[dlang-bot]

Thanks for your pull request, @ibuclaw!

Bugzilla references

Auto-close	Bugzilla	Severity	Description
✓	19912	critical	[module] No implicit import of object module when an object declaration exists.

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 "dmd-cxx + dmd#9999"

[JinShil]

Are you all sure about this? It seems that, instead, any object declaration would generate an error just like it does for TypeInfo See https://run.dlang.io/is/1mwpVi Don't get me wrong, I don't like the fact that these runtime objects must be declared in object.d, but it seems to be a more systemic issue than just the symbol object.

I'd prefer that object.d were not automatically imported at all, but I tried that, and was overruled.

Another option I'm considering is renaming object.d to package.d and have the compiler automatically import the runtime "package" instead of a specific module.

Can't we just have dmd.conf declare the automatic import of object.d or something so it's not hard-coded into the compiler like it currently is?

Anyway, can we discuss how to move forward on this? I think there's more to consider here than just what this PR addresses.

[thewilsonator]

I'd prefer that object.d were not automatically imported at all, but I tried that, and was overruled.

I think a better idea would be to have an auto imported package that would by default import object, that way we could split up the Frankenstein that is object.d. Then you could easily remove stuff that you don't want, since if you are going to the effort of not wanting classes deleting a file from the runtime that you are using is a lot easier than ripping out bits of object.d.

[JinShil]

It seems that, instead, any object declaration would generate an error just like it does for TypeInfo See https://run.dlang.io/is/1mwpVi

Oops, sorry. I see now that this PR is actually adding that behavior. That is consistent with the status quo, so 👍

I think this PR is OK as is, as it maintains the status quo, but there's still a systemic problem with the way D automatically imports the runtime and the fact that it requires certain symbols to be declared in a specific place. But, that's something that can be addressed later, I suppose.

[ibuclaw]

Yes, this just makes it so that you can't override the automatic import with another declaration of the same name.

[jacob-carlborg]

The error message mentions import, but there is no import in the source code. Can we come up with a better message?