Community
Participate
Working Groups
From the newsgroup: Hello, I believe I found a Java compiler bug in Eclipse M4 (my fifth or so ;-); see the method "badUninitializedVariableErrorMessage()" in the example below. /** * class CompileError */ public class CompileError { private static final class MethodProvider { // If you remove the 'throws Exception': The error Message disapears public Object openAppletJarOf(Object o) throws Exception { return null; } } /** * This method compiles with a Compiler Error: * "The local variable oResult may not have been initialized" * Question: Why? */ public static Object badUninitializedVariableErrorMessage(// ) throws Exception { final MethodProvider oMethodProvider = new MethodProvider(); final boolean bFlag1 = !true, bFlag2 = !true; // If you uncomment the next if-else: The error Message disapears if (!bFlag1) { final Object o2 = null; } else { final Object o2 = oMethodProvider.openAppletJarOf(null); } final Object oResult, o1 = null; if (o1 != null) { oResult = null; oMethodProvider.openAppletJarOf(null); } else { // If you uncomment the next line: The error Message disapears final Object o4 = oMethodProvider.openAppletJarOf(null); oResult = null; } return oResult; } } Kind Regards Mle
I confirm this is a bug. Will investigate.
Hipp Hipp, Hurra! If I report my 10. Compiler bug I want win a golden banana: Is that possible? :-}
*** Bug 45365 has been marked as a duplicate of this bug. ***
Sure ! <g> btw, I couldn't find your other 4 (or so) reported bugs against the compiler...
When I found bug 10 -- in one or two days ;-) -- I collect my other bug reports, most of them are reported by other people: I first ask in the newsgroup "Is it a bug", and most of the time a kind reader reports the bug for me.
Problem was due to a missing copy of a flow info during exception analysis. Thus a side-effect got performed on original object, affecting further diagnosis. Fixed. Regression test added InitializationTest#test169
/* Bug 6: I don't open a new bug report for this kind of bug, * because it is possible that you don't like my humor: * I don't want to waste a bug report for a (possibly) bad joke, * but a comment on an exisitng bug report ... . * If this kind of bug count, then the banana is near. */ /** * This method compiles with a Compiler Warning: * "The local variable bStopRecursion is never read" * Question: Why? * Possible fix: Change Warning to: * "The local variable bStopRecursion is superfluous" */ public static void usedVariableReportedAsUnread(final boolean bStop) { final boolean bStopRecursion; if (bStop) { // ok } else { usedVariableReportedAsUnread(bStopRecursion = true); } } Btw I can explain why this construct is useful ....
This warning is telling you that the variable is never read, thus could be optimized out. Note that for debugging purpose, one may choose to preserve these, then you should simply turn the warning off.
You should always enter a new bug for a new issue. We can close it if wrong, don't worry.
Continued from: https://bugs.eclipse.org/bugs/show_bug.cgi?id=45357 >This warning is telling you that the variable is never read, thus could be >optimized out. I think it is not clear that this variable is never read, it depends on the interpretation (it is clear that it can become optimzed away but that is another theme): * int a,b,c,d; * How to interprete: a=b=c=d; * Interprete as function call: * Interpretatin 1: =a(=(b,=(c,d))) // a,b,c,d never read * Interprete as decomposable expression: * Interpretatin 2: t1=(c=d); t2=(b=t1); t3=(a=t2) // a,b,c,d never read * Interprete as decomposable expression with reduced temp variable set: * Interpretatin 3: c=(c=d); b=(b=c); a=(a=b) // b,c,d read; a not * How to interprete: a=d; * Interpretation A: As atomic expression a=b; // a never read * Interpretation B: As minmal assignmend chain a=b; --> a= (a=b) // a read * My prefered interpretation of a=b; is the Interpretation B. >Note that for debugging purpose, one may choose to preserve these, >then you should simply turn the warning off. The is a bad workaround, because I always want a warning for really unused variables but my 'bStopRecursion' is used (in my interpretation B). Sorry, but you are my new victim for useless discussion (the last was Oliver Thomann a few month ago). Kind Regards
No problem. My point is more at the generated code level. There are 2 instructions for reading or writing a local variable. When stating 'a = b', variable b is read, and variable a is written. The warning is only indicating variable which are not read, not variables which aren't used at all (we originally implemented this, but user feedback conducted us to rather signal unread variables).
* I think that is not my point: * Assume there is a real piece of Java Code * int x; * Math.abs(x=1); * Your prefered way to byte code interpretes x as unread: * final int x; * // looks like raw SSA: static single assignment * final int t = x=1; // t is a Compiler internal variable * Math.abs(t); // read t not x * My prefered way to byte code reads x (instead of t) * final int x; * x=1; // Still real SSA but : * Math.abs(x); // x is read * I stop discussion, if get some food (perhaps a very small tangerine?)
Verified.
Backported to 2.1.x stream
Verified for 2.1.3 (M20040225)
/* this warning is right? */ final int a; System.out.println(a=1); // compiler sais cannot read 'a'
okano: please file separate bug report for your issue. Don't append to close ones.
