| Summary: | [DR147] g++ does not treat injected class name correctly. | ||
|---|---|---|---|
| Product: | gcc | Reporter: | Gabriel Dos Reis <gdr> |
| Component: | c++ | Assignee: | Jason Merrill <jason> |
| Status: | RESOLVED FIXED | ||
| Severity: | normal | CC: | andrew.stubbs, angray, chrbr, christian.joensson, dodji, dushistov, egallager, fang, gcc-bugs, giovannibajo, gp.bolton, hans.ekkehard.plesser, ian, in10semotions, marc.glisse, msebor, pi3orama, pinskia, Woebbeking |
| Priority: | P2 | Keywords: | accepts-invalid |
| Version: | 3.3.1 | ||
| Target Milestone: | 4.5.0 | ||
| See Also: |
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102156 https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102468 https://gcc.gnu.org/bugzilla/show_bug.cgi?id=34810 https://gcc.gnu.org/bugzilla/show_bug.cgi?id=33659 |
||
| Host: | Target: | *-*-* | |
| Build: | Known to work: | ||
| Known to fail: | Last reconfirmed: | 2005-12-11 23:25:28 | |
I can confirm this in the mainline (20030730). ICC 6.0 rejects this, so does EDG 2.40. Subject: Re: g++ does not treat injected class name correctly. "pinskia at physics dot uc dot edu" <gcc-bugzilla@gcc.gnu.org> writes: | I can confirm this in the mainline (20030730). | ICC 6.0 rejects this, so does EDG 2.40. It should be noted that the last two compilers behave as required by the standard. http://anubis.dkuug.dk/jtc1/sc22/wg21/docs/cwg_defects.html#147 -- Gaby *** Bug 16877 has been marked as a duplicate of this bug. *** A testcase is already in the tc1 testsuite (g++.dg/tc1/dr147.C), xfailed. I'm adding a cross-reference to this PR as we speak. *** Bug 15865 has been marked as a duplicate of this bug. *** *** Bug 11487 has been marked as a duplicate of this bug. *** Still accepts-invalid as of 4.2-20070430 (RC2). *** Bug 33659 has been marked as a duplicate of this bug. *** *** Bug 36306 has been marked as a duplicate of this bug. *** *** Bug 36501 has been marked as a duplicate of this bug. *** *** Bug 39924 has been marked as a duplicate of this bug. *** *** Bug 40294 has been marked as a duplicate of this bug. *** If you're sure that it's a bug why isn't it fixed yet? Is it that hard that you need more than six years? Sure it's no show stopper but it's annoying. I'm using -pedantic and -ansi to ensure platform independent code and then this... gcc is a free software project driven largely by volunteers. Interest in fixing accepts-invalid bugs is generally low; people are generally more interested in rejects-valid bugs, or in better optimizations, or in avoiding regressions. There are, unfortunately, many open bugs, many of which have higher priority than this one. If this bug is important to you, you do have the option of fixing it yourself. Ian, I know open source and I also know that some parts are more interesting than others :-) Most the time I'm a happy GCC user (sure, it could be faster but that's what compile farms are for). But this bug is assigned for nearly 6 years now that makes me a bit sad. *** Bug 42064 has been marked as a duplicate of this bug. *** Gaby, I'm sorry, are you actively working on this issue? Mine now. Subject: Bug 11764 Author: jason Date: Tue Nov 24 20:55:24 2009 New Revision: 154519 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=154519 Log: PR c++/42137 * parser.c (cp_parser_mem_initializer_id): Pass typename_type to cp_parser_class_name. (cp_parser_unqualified_id): Same, rather than class_type. PR c++/11764 * parser.c (cp_parser_expression_statement): Give helpful error for constructor name used as type. Modified: trunk/gcc/cp/ChangeLog trunk/gcc/cp/parser.c trunk/gcc/testsuite/ChangeLog trunk/gcc/testsuite/g++.dg/lookup/name-clash4.C trunk/gcc/testsuite/g++.dg/tc1/dr147.C trunk/gcc/testsuite/g++.old-deja/g++.pt/ctor2.C Fixed. *** Bug 43649 has been marked as a duplicate of this bug. *** Hi,
shouldn't
struct A
{
struct B {};
};
int main()
{
A::A::B b;
}
fail to compile? It compiles with 4.6 and 4.7.
Cheers,
André
No, that's valid. In the nested-name-specifier the second A only considers namespaces, types and templates whose specializations are types ([basic.lookup.qual]) so it doesn't name the constructor. OK, I read some explanations in the duplicates but still don't know why this should be valid code. IMO it makes no sense to write A:A:A:B and given the facts that 1) people don't write this intentionally 2) people expect this to be invalid (see number of duplicates) 3) some other compiler don't compile this couldn't you at least give us a warning (e. g. with -pedantic)? (In reply to comment #24) > OK, I read some explanations in the duplicates but still don't know why this > should be valid code. That is a question for the standards committee. > IMO it makes no sense to write A:A:A:B and given the facts that > 1) people don't write this intentionally > 2) people expect this to be invalid (see number of duplicates) And the cases that are invalid were fixed by my patch. > 3) some other compiler don't compile this Which? Recent versions of EDG and clang both accept the testcase in comment 22. > couldn't you at least give us a warning (e. g. with -pedantic)? -pedantic means "strictly enforce the standard"; since the standard says that the testcase in comment 22 is well-formed, -pedantic should not complain about it. Someone could add another warning flag to warn about this, but it isn't a priority for me. (In reply to comment #25) > (In reply to comment #24) > > 3) some other compiler don't compile this > > Which? Recent versions of EDG and clang both accept the testcase in comment > 22. MSVC 2010 In another duplicate ICC 6 was mentioned. > > couldn't you at least give us a warning (e. g. with -pedantic)? > > -pedantic means "strictly enforce the standard"; since the standard says that > the testcase in comment 22 is well-formed, -pedantic should not complain about > it. Someone could add another warning flag to warn about this, Yep, that is probably the better solution. (In reply to comment #26) > > Which? Recent versions of EDG and clang both accept the testcase in comment > > 22. > > MSVC 2010 "Other compilers have bugs" doesn't seem like a good reason to change GCC. > In another duplicate ICC 6 was mentioned. That's ancient, isn't it? |
cc1plus does not implement the class name injection correctly. In particular the snipped below should be rejected on the basis that A::A does not name a type (it designates a constructor) struct A { }; int main() { A::A a; // should be an ERROR }