Work On Issue 18172 - add scope to pointers in getopt to allow @safe

[JackStouffer]

No description provided.

[dlang-bot]

Thanks for your pull request, @JackStouffer!

Bugzilla references

Auto-close	Bugzilla	Severity	Description
✗	18172	enhancement	std.getopt should allow taking parameters by ref (like std.format.formattedRead), s.t. it can be used in @safe

[JackStouffer]

I'm not sure how we want to handle DIP1000 regressions, but we need a way to ensure things stay @safe

[wilzbach]

By #6278 and thus gradually enabling -dip1000?

[JackStouffer]

But this unittest needs to be annotated with @safe when -dip1000 is passed and @system when it's not to prevent any DIP1000 regressions. #6278 Doesn't have a mechanism for that.

[wilzbach]

Argh fair enough - there still might be issues as long as -deps compiles all unittests of its dependencies :/
Anyhow, for the time being #6278 (or similar could set DIP1000 by hand).

[JackStouffer]

this is due to the fact that to!(string)(string) acts as a pass through.

[dnadlinger]

.idup?

[wilzbach]

That really sucks. Safety shouldn't result in a performance overhead :/
Especially because in this case

• getoptImpl doesn't store anything, but modifies configuration and rslt
• it's a GC allocated string

[wilzbach]

Here's an idea how to programmatically detect -dip1000:

void main()
{
    enum isDIP1000 = __traits(compiles, () @safe {
         int x;
         int* p;
         p = &x;
    });
    pragma(msg, isDIP1000);
}

normal: https://run.dlang.io/is/RID7vh
-dip1000: https://run.dlang.io/is/1yJfVQ

Of course that could be fancier.

(based on the recent forum thread: https://forum.dlang.org/post/ayemvxctmpqqsokbmqeb@forum.dlang.org)

[JackStouffer]

If people want to tackle the DIP1000 returning a struct issue, I'd be glad to see their solution. Until then, I'm closing this as this isn't going to go anywhere.