C++ Curious case of passing typename/datatype name/class name to thread constructor

Question

I have returned to c++ after long years of Java/Node.js programming so, I am out of touch with the templates (I am working on that).

While revising concurrency, I accidentally stumbled upon the following case

class Test
{
};

int main()
{
  std::thread t(Test);
}

What I was experimenting with was, what if I pass a non-callable object and then I would study the error and source code. But I accidentally passed the type name. The program compiled an ran with no issue and I was baffled.

Then I tried something more stupid

int main()
{
  std::thread t1(int);
}

And somehow it worked as well. Though these two cases compiled and ran properly. I don't get how. If it tried t.join() or t1.join() I get following compilation error

t.join() cannot be resolved

I just simply just don't get what is happening in these two cases and how t.join() could not be resolved.

My research on it so far

I spend a significant amount of time trying to figure out how the compiler would have figured out whether the given value is callable. I came across this link: find out if a C++ object is callable It made a lot of sense because of member detection idiom but when I checked the source of gcc it seemed it used std::_Bind_simple of functional header/library to figure this out. But it didn't solve my query. Then, this post how std::thread constructor detects rvalue reference? asked the similar question but the answer made no sense to me.

Any sort of assistance will be very appreciated

Oh!! I have been victim to most vexing parse. If possible could you elaborate how the parsing is happing in this case. — Dr. Xperience, Nov 23 '17 at 17:30
@NeilButterworth: I won't say most vexing parse, as I don't see other way to parse it :) — Jarod42, Nov 23 '17 at 17:30
@Jarod42 I totally forgot about the declaration in local scope. I am really rusty. — Dr. Xperience, Nov 23 '17 at 17:31
I don't see why this is getting downvoted. It shows thought and effort and is expressed very clearly. — François Andrieux, Nov 23 '17 at 17:44
@FrançoisAndrieux Thanks for concern bro but the error is quite foolish. I am really quite embarrassed. — Dr. Xperience, Nov 23 '17 at 17:47
@Dr.Xperience: It is not foolish. I've been studying and using C++ for about 12 years now, and I still make this mistake every now and then. — Christian Hackl, Nov 23 '17 at 17:55
@Dr.Xperience It's still a good question. Remember that questions on Stack Overflow stay up for future readers. And you are probably not the only person to have made this mistake. Local function declarations are one of the great [gotcha](https://en.wikipedia.org/wiki/Gotcha_(programming))s of c++. — François Andrieux, Nov 23 '17 at 17:55
Oh, and I just realised that this is an especially devious case of the most vexing parse, because if you forget to `join` or `detach`, then you might easily be tricked into believing that a thread was actually started. — Christian Hackl, Nov 23 '17 at 17:57
@FrançoisAndrieux Hmm, I didn't thought. Good insight. Though net is very cruel. — Dr. Xperience, Nov 23 '17 at 17:57

score 3 · Accepted Answer · answered Nov 23 '17 at 17:39

Thanks to @Jaroda42 what I was doing was making a function declaration in local scope.

std::thread t(Test);

Where function being being t return type be std::thread and parameters being class type Test.

It was really stupid but this is what one get if one leaves a language for long time.

C++ Curious case of passing typename/datatype name/class name to thread constructor

1 Answers1