Bug 19656 – D compiler fails to resolve circular module dependency when modules are compiled separately

Created attachment 1730 Tar Gzipped code Tar gzipped file is attached. To reproduce the bug, untar and compile foo.d. Note that when all files/modules are compiled together, the compiler does not give any error. This is what I get: $ dmd -c foo.d fred.d(4): Error: function void fred.qux!(bar).qux.thunk() does not override any function, did you mean to override void foo.foo.thunk()? fred.d(7): Error: template instance `fred.qux!(bar)` error instantiating

Also see related bug https://issues.dlang.org/show_bug.cgi?id=19657

Works with version 2.074.1 of DMD. Fails with 2.075.0 and later. Marking as regression.

Any chance you could reduce this further? (Same also for the other bugs that you opened). This would help a lot for debugging them.

I will shortly add a PR for this and for other similar issues.

@puneet created dlang/dmd pull request #9462 "Fix for issue 19656" mentioning this issue: - Fix for issue 19656 When DMD encounters an associative array with a class type (say Foo) specified as key, the compiler immediately wants to analyze the Foo class. And in some cases it might so happen that DMD was already in the process of analyzing Foo. In such a scenario, the visit(ClassDeclaration ) method malfunctions and marks Foo with PASS.semanticdone without actually completing analysis of the class. Since this method marks a ongoing call with setting _scope to null, this situation can be avoided by testing _scope for Foo before calling the dsymbolSemantic method. This makes sure that if semantic analysis for Foo is already running, it would not be run again. https://github.com/dlang/dmd/pull/9462

@puneet created dlang/dmd pull request #9471 "Fix issues 19655, 19656, 19657, 19746 & 19750" fixing this issue: - Fix issues 19655, 19656, 19657, 19746 & 19750 The Semantic (pass 1) analysis for classes is handled by visit(ClassDeclaration ) method of DsymbolSemanticVisitor class. For a given class, this method may be run multiple times in order to resolve forward references. The method incrementally tries to resolve the types referred to by the members of the class. The subsequent calls to this method are short-circuited if the class members have been fully analyzed. For this the code tests that it is not the first/main call to the method (semanticRun == PASS.init else branch), scx is not set, and that the cldec.symtab is already set. If all these conditions are met, the method returns. But before returning, the method was setting cldec.semanticRun to PASS.semanticdone. It should not set semanticRun since the class has not been fully analyzed yet. The base class analysis for this class could be pending and as a result vtable may not have been fully created. This fake setting of semanticRun results in the semantic analyzer to believe that the class has been fully analyzed. As exposed by the issues 19656, 19657, 19746 and 19750, it may result in compile time errors when a derived type class is getting analyzed and because of this fake semanticdone on the base class, the semantic analysis construes that an overridden method is not defined in the base class. Issue 19655 exposes a scenario where a buggy vtable may be created and a call to class method may result in execution of some adhoc code. https://github.com/dlang/dmd/pull/9471

dlang/dmd pull request #9471 "Fix issues 19655, 19656, 19657, 19746 & 19750" was merged into stable: - 592d0c790b90be5aab9d011200b98efa539b4a27 by Puneet Goel: Fix issues 19655, 19656, 19657, 19746 & 19750 The Semantic (pass 1) analysis for classes is handled by visit(ClassDeclaration ) method of DsymbolSemanticVisitor class. For a given class, this method may be run multiple times in order to resolve forward references. The method incrementally tries to resolve the types referred to by the members of the class. The subsequent calls to this method are short-circuited if the class members have been fully analyzed. For this the code tests that it is not the first/main call to the method (semanticRun == PASS.init else branch), scx is not set, and that the cldec.symtab is already set. If all these conditions are met, the method returns. But before returning, the method was setting cldec.semanticRun to PASS.semanticdone. It should not set semanticRun since the class has not been fully analyzed yet. The base class analysis for this class could be pending and as a result vtable may not have been fully created. This fake setting of semanticRun results in the semantic analyzer to believe that the class has been fully analyzed. As exposed by the issues 19656, 19657, 19746 and 19750, it may result in compile time errors when a derived type class is getting analyzed and because of this fake semanticdone on the base class, the semantic analysis construes that an overridden method is not defined in the base class. Issue 19655 exposes a scenario where a buggy vtable may be created and a call to class method may result in execution of some adhoc code. https://github.com/dlang/dmd/pull/9471

dlang/dmd pull request #9489 "Merge remote-tracking branch 'upstream/stable' into merge_stable" was merged into master: - 58878aeb8d32474ac24a0dd51446533a93602564 by Puneet Goel: Fix issues 19655, 19656, 19657, 19746 & 19750 (#9471) Fix issues 19655, 19656, 19657, 19746 & 19750 merged-on-behalf-of: Nicholas Wilson <[email protected]> https://github.com/dlang/dmd/pull/9489

dlang/dmd pull request #10913 "[dmd-cxx] Fix issues 19655, 19656, 19657, 19746 & 19750 (#9471)" was merged into dmd-cxx: - 968386f6da15cb30c229bf0ca3325d86c6fa8714 by Puneet Goel: Fix issues 19655, 19656, 19657, 19746 & 19750 (#9471) Fix issues 19655, 19656, 19657, 19746 & 19750 merged-on-behalf-of: Nicholas Wilson <[email protected]> https://github.com/dlang/dmd/pull/10913