Fix Issue 17382 - void main(){}pragma(msg,main()); crashes DMD

[RazvanN7]

No description provided.

[dlang-bot]

Thanks for your pull request, @RazvanN7! We are looking forward to reviewing it, and you should be hearing from a maintainer soon.

Some tips to help speed things up:

• smaller, focused PRs are easier to review than big ones

• try not to mix up refactoring or style changes with bug fixes or feature enhancements

• provide helpful commit messages explaining the rationale behind each change

Bear in mind that large or tricky changes may require multiple rounds of review and revision.

Please see CONTRIBUTING.md for more information.

Bugzilla references

Auto-close	Bugzilla	Description
✓	17382	void main(){}pragma(msg,main()); crashes DMD

[stefan-koch-sociomantic]

NO!
That'll generate an uncompilable version.
As well as silencing real errors!
I suggest you look for a different fix.

[RazvanN7]

NO!
That'll generate an uncompilable version.

Can you be more specific?

[stefan-koch-sociomantic]

the generated header will have <void> in it which has no place in our langauge.
Also If the case happens which would have asserted before it will now get swallowed.

[RazvanN7]

You might be right, however asserts are not errors; assert(0) is used on code paths that should never be taken and the user should never be faced with a failed assert so from this point of view no error is swollen.

On the other hand, if you use pragma(msg, foo()) where foo is declared as void foo() {} the output is going to be < void > . Now, in the case of this PR, I agree that there might be cases where < void > is outputted and it shouldn't but I don't know any and it seems that the tests all pass. So, when I was asking you to be specific I was expecting an example. Thank you.

[stefan-koch-sociomantic]

I cannot think of an example as of now.
As the assert would have triggered in such a case.
However that is no excuse to open us up to more subtle bugs.

Besides having the hdrgen write-out invalid D is hardly an acceptable solution.

I'd rather fix this at the point where the main-symbol is rewritten.

[MetaLang]

Excuse the drive-by comment given that I know nothing about the compiler, but why does the compiler even try to const-fold main?

[andralex]

It doesn't matter that it's called main, does it? @RazvanN7 please arrange that a compile-time error with a good message is issued if someone attempts to pragma msg a void expression. If anyone does want to, there are many means such as pragma(msg, (main(), "")); etc

[RazvanN7]

@andralex Is this ok?

[nemanja-boric-sociomantic]

Missing () at toChars call (style doesn't match anywhere in this module).

[andralex]

"has type void" is more pedantically correct, but this is fine as is - thanks