www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.learn - Possible bug in RVO?

reply Yuxuan Shui <yshuiv7 gmail.com> writes:
I have encountered a weird bug.

I defined a Set class, which has a opBinary!"-". And somehow this:

auto tmp = set_a-set_b;

produces different results as this:

set_a = set_a-set_b;

the latter will produce an empty set.

I tried to reduce the source code to get a test case. But this 
problem just goes away after removing some code.

Any ideas what I could have done wrong?
Apr 03 2016
next sibling parent Yuxuan Shui <yshuiv7 gmail.com> writes:
On Monday, 4 April 2016 at 03:28:01 UTC, Yuxuan Shui wrote:

 I have encountered a weird bug.

 I defined a Set class, which has a opBinary!"-". And somehow 
 this:

 auto tmp = set_a-set_b;

 produces different results as this:

 set_a = set_a-set_b;

 the latter will produce an empty set.

 I tried to reduce the source code to get a test case. But this 
 problem just goes away after removing some code.

 Any ideas what I could have done wrong?


OK, I think I got a test case:

import std.traits;
struct Set {
public:
	void insert(ulong v) {
		aa[v] = true;
	}
	size_t size() const {
		return aa.length;
	}
	auto opBinary(string op)(ref Set o) const {
		Set ret;
		foreach(k; aa.byKey)
			if (k !in o.aa)
				ret.insert(k);
		return ret;
	}
	 disable this(this);
	bool[ulong] aa;
}

struct XX {
	Set a, b, tmp;
	this(int n) {
		a.insert(n);
		tmp = a-b;
		a = a-b;
	}
}
void main(){
	import std.stdio;
	XX xx = XX(1000);
	writeln(xx.a.size);
	writeln(xx.tmp.size);
}

This does not happen when 'a' is on stack, that's why I was 
having trouble reproducing it.

I don't think this is valid code, because Set has disabled 
post-blit, 'a = a-b' should report an error. However, I don't 
think current behavior of dmd is correct either.
Apr 03 2016
prev sibling parent reply Yuxuan Shui <yshuiv7 gmail.com> writes:
On Monday, 4 April 2016 at 03:28:01 UTC, Yuxuan Shui wrote:

 I have encountered a weird bug.

 I defined a Set class, which has a opBinary!"-". And somehow 
 this:

 auto tmp = set_a-set_b;

 produces different results as this:

 set_a = set_a-set_b;

 the latter will produce an empty set.

 I tried to reduce the source code to get a test case. But this 
 problem just goes away after removing some code.

 Any ideas what I could have done wrong?


A slightly more reduced test case:

struct Set {
	void insert(ulong v) {
		aa[v] = true;
	}
	 disable this(this);
	bool[ulong] aa;
}
auto clobber(ref Set x, ref Set o) {
	Set ret;
	ret.aa = x.aa;
	return ret;
}
struct XX {
	Set a, b, tmp;
	this(int n) {
		a.insert(1);
		//a.aa[1] = true;  <--- Swap above line with this doesn't 
trigger the bug
		tmp = a.clobber(b);
		a = a.clobber(b);
	}
}
void main(){
	import std.stdio;
	XX xx = XX(0);
	writeln(xx.a.aa.length);
	writeln(xx.tmp.aa.length);
}
Apr 03 2016
next sibling parent Yuxuan Shui <yshuiv7 gmail.com> writes:
On Monday, 4 April 2016 at 03:55:26 UTC, Yuxuan Shui wrote:

 On Monday, 4 April 2016 at 03:28:01 UTC, Yuxuan Shui wrote:

 I have encountered a weird bug.

 I defined a Set class, which has a opBinary!"-". And somehow 
 this:

 auto tmp = set_a-set_b;

 produces different results as this:

 set_a = set_a-set_b;

 the latter will produce an empty set.

 I tried to reduce the source code to get a test case. But this 
 problem just goes away after removing some code.

 Any ideas what I could have done wrong?


 A slightly more reduced test case:


And LDC has the same problem with the first test case, but not 
with the second one.
Apr 03 2016
prev sibling parent reply Anonymouse <asdf asdf.com> writes:
On Monday, 4 April 2016 at 03:55:26 UTC, Yuxuan Shui wrote:

 On Monday, 4 April 2016 at 03:28:01 UTC, Yuxuan Shui wrote:
 auto clobber(ref Set x, ref Set o) {
 	Set ret;
 	ret.aa = x.aa;


  	  assert(x.aa.length > 0);  // <-- boom

 	return ret;
 }


No idea myself but that's where it seems to go wrong.
Apr 04 2016
parent reply =?UTF-8?Q?Ali_=c3=87ehreli?= <acehreli yahoo.com> writes:
On 04/04/2016 09:36 AM, Anonymouse wrote:

 On Monday, 4 April 2016 at 03:55:26 UTC, Yuxuan Shui wrote:

 On Monday, 4 April 2016 at 03:28:01 UTC, Yuxuan Shui wrote:
 auto clobber(ref Set x, ref Set o) {
     Set ret;
     ret.aa = x.aa;


         assert(x.aa.length > 0);  // <-- boom

     return ret;
 }


 No idea myself but that's where it seems to go wrong.


Looks like a bug. Just to make it more visible:

auto clobber(ref Set x, ref Set o) {
     writefln("entered clobber    - x.aa: %s", x.aa);
     Set ret;
     writefln("did not touch x.aa - x.aa: %s", x.aa);
     ret.aa = x.aa;
     return ret;
}

x.aa changes during the second call:

entered clobber    - x.aa: [1:true]
did not touch x.aa - x.aa: [1:true]
entered clobber    - x.aa: [1:true]
did not touch x.aa - x.aa: []          <-- What happened?

Ali
Apr 04 2016
parent Yuxuan Shui <yshuiv7 gmail.com> writes:
On Monday, 4 April 2016 at 21:31:08 UTC, Ali Çehreli wrote:

 On 04/04/2016 09:36 AM, Anonymouse wrote:

 On Monday, 4 April 2016 at 03:55:26 UTC, Yuxuan Shui wrote:

     [...]


         assert(x.aa.length > 0);  // <-- boom

 [...]


 No idea myself but that's where it seems to go wrong.


 Looks like a bug. Just to make it more visible:

 auto clobber(ref Set x, ref Set o) {
     writefln("entered clobber    - x.aa: %s", x.aa);
     Set ret;
     writefln("did not touch x.aa - x.aa: %s", x.aa);
     ret.aa = x.aa;
     return ret;
 }

 x.aa changes during the second call:

 entered clobber    - x.aa: [1:true]
 did not touch x.aa - x.aa: [1:true]
 entered clobber    - x.aa: [1:true]
 did not touch x.aa - x.aa: []          <-- What happened?

 Ali


I filed an bug report here: 
https://issues.dlang.org/show_bug.cgi?id=15869
Apr 04 2016