Weird use of '...' in function argument

[ignatirabo]

When decompiling a program, I have found many functions that have the following signature.

However, when I inspect the signature of the function I see this:

As soon as I press TAB it changes to void:

And now the function arguments is just void (as I would expect).

Is this expected behaviour? What does void... even mean? Is it just void? If it is void why have the ... extra and not just void?

I am using Ghidra I compiled myself based on commit 79de6c6bf938b9c85c0cb6ce38043ea0f5819b6c

• OS: Ubuntu 24.10
• Java Version: Amazon Corretto 21
• Ghidra Version: 11.4-DEV
• Ghidra Origin: Ghidra I compiled myself based on commit 79de6c6bf938b9c85c0cb6ce38043ea0f5819b6c

[gemesa]

That is how varargs (variable arguments) work. You even have it enabled in your first picture. When you hit TAB, you disable it.

[ignatirabo]

Ok I see, so it's just the printing that threw me off with void....
Other functions with variable arguments usually look like: void f(int a, ...) and not void f(int a, void...).

If I made a change to Ghidra to print it properly, would the Ghidra devs be interested?

[gemesa]

You need to uncheck varargs, set the parameter type (e.g. to int), check varargs and then it will be displayed properly:

[dgutson]

The point here is that the generated code syntax is not C valid.

[gemesa]

Well, what code should the decompiler show for nonsense inputs like a parameterless (void) function which uses varargs at the same time?

[dgutson]

then the question is, why Ghidra concludes that the function is parameterless and uses var args in the first place? I'm not being rhetoric, I am really asking.

[gemesa]

It makes assumptions which are sometimes correct and sometimes wrong. In the incorrect cases you need to fix it and help the decompiler. Since void and varargs together do not make sense, you can open a bug report, attach your binary and see what the Ghidra team says.