Bug 209 – "Diamond" import name conflicts when using Fully Qualified names

In a "diamond-shaped" structure of imports, name conflicts occur on the bottom module when accessing top module entities with a fully qualified name (because the top module name itself conflicts). Example: ---- foo.d ---- module foo; int fooEnt; // could a be a class, func, any other entity. ---- barA.d ---- module barA; import foo; ---- barB.d ---- module barA; import foo; ---- test.d ---- import barA; import barB; void main() { fooEnt++; // Compiles ok. (Correct) foo.fooEnt++; // Compile error, name conflicts (INCORRECT) : // barA.d(3): import barA.foo conflicts with barB.foo at barB.d(3) alias foo xxx; // Compile error too, // accessing the top module is enough to trigger it } This bug does not happen if the "root name" (first name) of the top and middle modules is the same, such as this: ---- mod.foo.d ---- module mod.foo; int fooEnt; // could a be a class, func, any other entity. ---- mod.barA.d ---- module mod.barA; import mod.foo; ---- mod.barB.d ---- module mod.barA; import mod.foo; ---- test.d ---- import mod.barA; import mod.barB; void main() { fooEnt++; // Compiles ok. (Correct) mod.foo.fooEnt++; // Compiles ok. (Correct) }

(In reply to comment #0) > ---- barB.d ---- > module barA; > > import foo; Shouldn't this be module barB; ?

(In reply to comment #1) > (In reply to comment #0) > > ---- barB.d ---- > > module barA; > > > > import foo; > Shouldn't this be > module barB; > ? Agh, yes indeed. :/ Same for: ---- mod.barB.d ---- module mod.barA; // shoulde be barB

Hum, seems this was partially fixed in DMD.161 : The example above works since DMD.161, but there is one other that doesn't currently(.162), this one is not "diamond" shaped anymore since the top is folded into two. Let's see the code: ---- test.d ---- module test; import tierOneA; import tierOneB; alias pack.fooA XXX; // the above gets the error: // import tierOneA.pack conflicts with tierOneB.pack at tierOneB.d(3) ---- tierOneA.d ---- module tierOneA; import pack.fooA; ---- tierOneB.d ---- module tierOneB; import pack.fooB; ---- pack.fooA ---- module pack.fooA; // this is the second tier A ---- pack.fooB ---- module pack.fooB; // this is the second tier B

(In reply to comment #3) > ... Add "public" to the imports of tierOneA and tierOneB to update this example to DMD 0.163 . (Although the error still appears with private imports)

New bug cloned, this one (mysteriously?) fixed on DMD.161 .

For the audit log, and in case some sort of data inconsistency arises, I performed these sql commands to delete attachment and comment spam: mysql> delete from attachments where attach_id >= 107 and attach_id <= 136; Query OK, 30 rows affected (0.00 sec) mysql> delete from longdescs where bug_id=209 and who=223; Query OK, 30 rows affected (0.00 sec)