Object#const_defined? triggers a warning for deprecated constant #5002
Comments
janko
changed the title
|
Ok I'm really confused why MRI does not display this warning. As far as I can tell, the const_defined? code calls rb_const_get which calls rb_const_warn_if_deprecated, which always warns. I'm missing something. |
|
Ok, I figured it out by step debugging. If However we do not appear to have that simpler logic. RubyModule.isConstantDefined is similar but not quite. |
|
I am generally in favor of matching MRI behavior but this really appears to be a bug in MRI. Why would "Kernel" as argument trigger it while :Kernel does not? I think we should WONTFIX this and perhaps ask @janko-m to raise this on ruby-lang as an issue. |
|
@enebo Ahh you are correct, that does seem to be an odd behavioral difference. I'd say that all forms of @janko-m I would suggest opening a bug with MRI to clarify this behavior. I did try to port the relevant logic, and although it eliminates the warning it does not pass existing specs (so I missed something). If MRI decides that this is correct behavior we can proceed to use this patch as a starting point to getting this special case to work the same way. diff --git a/core/src/main/java/org/jruby/RubyModule.java b/core/src/main/java/org/jruby/RubyModule.java
index 59ea1b9ece..81cda9728a 100644
--- a/core/src/main/java/org/jruby/RubyModule.java
+++ b/core/src/main/java/org/jruby/RubyModule.java
@@ -3245,6 +3245,10 @@ public class RubyModule extends RubyObject {
throw runtime.newNameError("wrong constant name ", fullName);
}
+ if (sep == -1 && symbol instanceof RubySymbol) {
+ return runtime.newBoolean(mod.rb_const_defined_0(runtime, name, false, true, false));
+ }
+
IRubyObject obj;
while ( ( sep = name.indexOf("::") ) != -1 ) {
final String segment = name.substring(0, sep);
@@ -3265,6 +3269,39 @@ public class RubyModule extends RubyObject {
return runtime.newBoolean(obj != null);
}
+ private boolean
+ rb_const_defined_0(Ruby runtime, String id, boolean exclude, boolean recurse, boolean visibility)
+ {
+ RubyModule tmp;
+ boolean mod_retry = false;
+ ConstantEntry ce;
+
+ tmp = this;
+ retry: while (true) {
+ while (tmp != null) {
+ if ((ce = tmp.constantEntryFetch(id)) != null) {
+ if (visibility && ce.hidden) {
+ return false;
+ }
+ if (ce.value == UNDEF && //!check_autoload_required(tmp, id, 0) &&
+ tmp.getAutoloadConstant(id, false) == null)
+ return false;
+ return false;
+ }
+ if (!recurse) break;
+ tmp = tmp.superClass;
+ }
+ if (!exclude && !mod_retry && this.metaClass == runtime.getModule()) {
+ mod_retry = true;
+ tmp = runtime.getObject();
+ continue retry;
+ }
+ break retry;
+ }
+ return false;
+ }
+
+
/** rb_mod_const_get
*
*/ |
|
@headius Thanks for the explanation, I will take it to the MRI team. |
Environment
Expected Behavior
After deprecating a constant, asking whether that constant is defined doesn't trigger a deprecation warning on MRI:
Actual Behavior
On JRuby the
Object#const_defined?call triggers a deprecation warning:I'm not saying this behaviour is bad, but it would probably be nice if the behaviour was the same as MRI when possible.
The text was updated successfully, but these errors were encountered: