On Mon, Apr 18, 2011 at 1:40 PM, Ulf Zibis <ulf.zi...@gmx.de> wrote: > Am 15.04.2011 15:43, schrieb Michael McMahon: >> >> I have incorporated much of Ulf's approach into this version. >> >> I agree with Alan about the spec. We could find other variables in the >> future that need >> to be set, but can have alternative values other than the default. I think >> this approach >> is the most flexible. >> >> http://cr.openjdk.java.net/~michaelm/7034570/webrev.4/ >> > > Additionally I see another problem: > > Because the ProcessEnvironment.nameComparator is static, creating a new > Process will not be thread-safe!
Ulf, I don't see a problem here. The ProcessEnvironment.nameComparator is a static final field that is initialized to `new NameComparator()` in a static initialization block for the ProcessEnvironment class. The NameComparator has no instance state, and can be safely shared by multiple threads. >From >http://cr.openjdk.java.net/~michaelm/7034570/webrev.4/src/windows/classes/java/lang/ProcessEnvironment.java.html: 191 private static final class NameComparator 192 implements Comparator<String> { 193 public int compare(String s1, String s2) { 194 // We can't use String.compareToIgnoreCase since it 195 // canonicalizes to lower case, while Windows 196 // canonicalizes to upper case! For example, "_" should 197 // sort *after* "Z", not before. 198 int n1 = s1.length(); 199 int n2 = s2.length(); 200 int min = Math.min(n1, n2); 201 for (int i = 0; i < min; i++) { 202 char c1 = s1.charAt(i); 203 char c2 = s2.charAt(i); 204 if (c1 != c2) { 205 c1 = Character.toUpperCase(c1); 206 c2 = Character.toUpperCase(c2); 207 if (c1 != c2) 208 // No overflow because of numeric promotion 209 return c1 - c2; 210 } 211 } 212 return n1 - n2; 213 } 214 } ... 227 private static final NameComparator nameComparator; ... 233 static { 234 nameComparator = new NameComparator(); ... - Dave
