www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.bugs - strange seg-v on reclusive call

reply BCS <ao pathlink.com> writes:
This looks to me like a bug but I could be wrong.

private import std.stdio;

void Part(Rule[] inSet)
{
	scope(failure) writef("===========\n");

	
	{ // dorp this blobk and things work
		Build!(Rule) tmp;

		insertUnUsed: foreach(r; inSet)
			if(r.users.length == 0)	// drop this line and it runs
			{
			}
	}

	F(   [new Rule, null, null, null, null]);	// this works

	scope(failure) writef("+++++++++++++++\n");

	Part([new Rule, null, null, null, null]);	// this segv's

	scope(failure) writef("----------\n");
}

void F(Rule[] r){}

class Rule { Rule[] users; }

alias void function(int) sighandler_t;
extern (C) sighandler_t signal(int signum, sighandler_t handler);

void main()
{

	// convert sigsegv to excption
	signal(11,function void(int i){throw new Error("SEGV");});

	Part([new Rule()]);
}


struct Build(T)
{
	static const int step = 5;
	T[] a;
	int at = 0;

	void Insert(T t)
	{
		if(a.length <= at)
			a.length = at + step;
		a[at++] = t;
	}
}
Sep 17 2007
parent BCS <ao pathlink.com> writes:
Reply to Benjamin,

this is definitely a code gen bug running with -O fixes it (on the current 
version). By looking at the ASM the only difference seems to be in the register 
assignments, possibly indicating an unassigned value bug (there is a register 
that is used without being assigned after that last statement).

I'm posting a bug.
Sep 18 2007