| Summary: | [3.3/3.4 regression] template member function (getting a nested template as parameter) cannot be called anymore if another unrelated template member function is defined. | ||
|---|---|---|---|
| Product: | gcc | Reporter: | have |
| Component: | c++ | Assignee: | Not yet assigned to anyone <unassigned> |
| Status: | RESOLVED FIXED | ||
| Severity: | normal | CC: | gcc-bugs, giovannibajo |
| Priority: | P3 | Keywords: | rejects-valid |
| Version: | 3.3 | ||
| Target Milestone: | 3.3.1 | ||
| Host: | Target: | ||
| Build: | Known to work: | ||
| Known to fail: | Last reconfirmed: | ||
State-Changed-From-To: open->analyzed
State-Changed-Why: Confirmed. The funny thing about this bug is that if you
remove the member functions g() and h(), the call to f()
succeeds.
I rate this as a regression because the above snippet
without h() compiles with 2.95 but fails on 3.x. The snippet
provided by the poster fails also with 2.95, but it's the
opposite problem: it manages to call f() but not g() and
h().
Subject: Bug 10845 CVSROOT: /cvs/gcc Module name: gcc Changes by: mmitchel@gcc.gnu.org 2003-06-20 15:44:25 Modified files: gcc/cp : ChangeLog pt.c gcc/testsuite : ChangeLog Added files: gcc/testsuite/g++.dg/template: member3.C Log message: PR c++/10845 * pt.c (try_class_unification): Correct handling of member class templates. * semantics.c (genrtl_finish_function): Adjust expand_function_end call. Patches: http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/ChangeLog.diff?cvsroot=gcc&r1=1.3443&r2=1.3444 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/pt.c.diff?cvsroot=gcc&r1=1.703&r2=1.704 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&r1=1.2784&r2=1.2785 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/g++.dg/template/member3.C.diff?cvsroot=gcc&r1=NONE&r2=1.1 Subject: Bug 10845 CVSROOT: /cvs/gcc Module name: gcc Branch: gcc-3_3-branch Changes by: mmitchel@gcc.gnu.org 2003-06-20 15:46:42 Modified files: gcc/cp : ChangeLog pt.c gcc/testsuite : ChangeLog Added files: gcc/testsuite/g++.dg/template: member3.C Log message: PR c++/10845 * pt.c (try_class_unification): Correct handling of member class templates. PR c++/10845 * g++.dg/template/member3.C: New test. Patches: http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=1.3076.2.156&r2=1.3076.2.157 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/pt.c.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=1.635.2.29&r2=1.635.2.30 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=1.2261.2.194&r2=1.2261.2.195 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/g++.dg/template/member3.C.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=NONE&r2=1.1.2.1 Fixed in GCC 3.3.1, GCC 3.4. |
error: no matching function for call to `A<int>::f(int&, A<int>::SubA<int>&)'. In what follows, words written as _underscored_ mean that without any _underscored_ hypothesis, bug does not occur. Let us consider a method f (with _two_ templates arguments) of a _template_ class A. The first argument is given by a _template T1_. The second argument is given as a _template_ sub class of A (SubA<L>) and the second template typename of f is given for matching L. Moreover, there exists (at least) another template method in A which has the same argument type than the second one of f. Then f cannot be found when it is called. Release: g++-3.2, 3.3 Environment: Linux i686 How-To-Repeat: Try to compile this code ---8<------------ template<typename T> struct A { template<typename L> struct SubA { }; template<typename T1,typename L> void f(T1 & t1, SubA<L> & t2) { } template<typename U> void g(SubA<U> & suba) { } template<typename U> void h(SubA<U> & suba) { } }; int main(void) { int i; A<int> a; A<int>::SubA<int> suba; a.f(i,suba); a.g(suba); a.h(suba); }