www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.bugs - Bug in string or AA handling

reply =?ISO-8859-1?Q?Jari-Matti_M=E4kel=E4?= <jmjmak utu.fi.invalid> writes:
This code segfaults on linux with dmd 0.150 even with all debug-flags on:

import std.string, std.date, std.random;
alias char[] String;

class Node {
    String label;
    Node[char] successors;

    this(String l) { label = l; }
    void countCommonPrefix(String str) {
        int a = str.length, b = label.length, m = a < b ? a : b, c;
        while (c < m && str[c] == label[c]) c++;
    }
    void add(String str) {
        if (!(str[0] in successors)) successors[str[0]] = new Node(str);
        successors[str[0]].countCommonPrefix(str);
    }
}

void main() {
    String data;
    for(int i=0; i<100; i++)
        data ~= lowercase[rand()%lowercase.length];

    real estimate = getUTCtime() + TicksPerSecond*10;
    while (getUTCtime<estimate) {
        auto root = new Node("a");
	for (int i = 0; i < data.length; ++i) root.add(data[i..$]);
    }
}

Any idea, why it does so? If you shorten the time interval or data
string length, it may not segfault. This is a "minimal" test case made
of the suffix tree algorithm I had previously problems with. Another way
to prevent segfaults may (YMMV) be to use this "fix":

row 11:
-       while (c < m && str[c] == label[c]) c++;
+       while (c < m) if (str[c]==label[c]) c++;

I don't understand - &&-expressions should actually work this way. See
http://www.digitalmars.com/d/expression.html#AndAndExpression.

-- 
Jari-Matti
Mar 31 2006
parent reply Walter Bright <newshound digitalmars.com> writes:
Jari-Matti Mäkelä wrote:
 This code segfaults on linux with dmd 0.150 even with all debug-flags on:
 
 import std.string, std.date, std.random;
 alias char[] String;
 
 class Node {
     String label;
     Node[char] successors;

It does look like the AA problem with the key type not being a multiple size of a pointer size. A fix for this will go out in the next update.
Mar 31 2006
parent =?ISO-8859-1?Q?Jari-Matti_M=E4kel=E4?= <jmjmak utu.fi.invalid> writes:
Walter Bright wrote:
 Jari-Matti Mäkelä wrote:
 This code segfaults on linux with dmd 0.150 even with all debug-flags on:

 import std.string, std.date, std.random;
 alias char[] String;

 class Node {
     String label;
     Node[char] successors;

It does look like the AA problem with the key type not being a multiple size of a pointer size. A fix for this will go out in the next update.

Ok, thanks. I knew there was an AA thing already in the bugzilla, but you never can be too sure :)
Mar 31 2006