Skip to content

Conversation

@ThadHouse
Copy link
Contributor

Previously, all conditionals were marked as not constant. However, if the Cond, LHS and RHS are all constant, the ternary can be marked constant as well

Previously, all conditionals were marked as not constant. However, if the Cond, LHS and RHS are all constant, the ternary can be marked constant as well
}}
}}
";
var diagnostics = new Diagnostic[] { new Diagnostic(DiagnosticLevel.Warning, "Function like macro definition records are not supported: 'TESTRESULT_FROM_WIN32'. Generated bindings may be incomplete.", "Line 2, Column 9 in ClangUnsavedFile.h") };
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is there a better way to do this?

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not really, this is about how everything works for validating diagnostics.

{
return false;
var conditionalOperator = (ConditionalOperator)initExpr;
return IsConstant(conditionalOperator.Cond) && IsConstant(conditionalOperator.LHS) && IsConstant(conditionalOperator.RHS);
Copy link
Member

@tannergooding tannergooding Mar 28, 2021

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: This isn't quite right. If IsConstant(conditionalOperator.Cond) is true and the evaluated value of Cond is true, then only IsConstant(conditionalOperator.LHS) needs to be true.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

a ? b :c does that mean Cond is a and LHS is b : c? Otherwise confused by why only 2 of the 3 would need to be true.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This code doesn't check that Cond evaluates to true. It just checks that it evaluates to a constant.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

cond ? LHS : RHS

x = cond ? LHS : RHS is basically shorthand for:

if (cond)
{
    x = LHS;
}
else
{
    x = RHS;
}

So if you have x = trueCond ? 5 : sqrt(globalValue), then you want it to still evaluate to be "constant". As it is now, it won't since IsConstant(conditionalOperator.RHS) will return false

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Likewise if you had x = falseCond ? sqrt(globalValue) : 5, you want it to evaluate as constant while IsConstant(conditionalOperator.LHS) will be the only thing to return false

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Since the generator actually generates the ternary even if the Cond is constant, wouldn't the current impl need both LHS and RHS to be true? I agree if the generator actually lowered the ternary then only the evaluated condition would need to be checked, but thats not what gets generated currently.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks like yes actually. I thought C# would drop the dead code path here, but looks like its not.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah. I agree if the generator is changed to lower everything, then this evaluation could be more fine grained. But for now, it will generate the whole thing, so all 3 need to be constant evaluated in order to generate const.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks, this looks good and I'll merge once CI finishes

@ThadHouse
Copy link
Contributor Author

Anything we can do to poke the windows build thats been stuck for a few weeks?

@tannergooding
Copy link
Member

It actually has passed and the status never got updated on the GitHub side. I'll force merge.

@tannergooding tannergooding merged commit cbe1fb3 into dotnet:main Apr 8, 2021
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants