This is GCC Bugzilla
This is GCC Bugzilla Version 2.20+
View Bug Activity | Format For Printing | Clone This Bug
The code a.C is rejected by gcc 3.4. This is a regression with respect to gcc >3.3. Environment: System: Linux linux 2.4.20-4GB #1 Fri Jul 11 07:33:18 UTC 2003 i686 unknown unknown GNU/Linux Architecture: i686 SuSE 8.2 host: i686-pc-linux-gnu build: i686-pc-linux-gnu target: i686-pc-linux-gnu configured with: ../gcc/configure --enable-threads=posix --enable-languages=c,c++,f77,objc --enable-__cxa_atexit --enable-debug How-To-Repeat: source code a.C #include <stdarg.h> class A { public: A(); A(const A& x); A& operator= (const A& a); ~A(); }; void f(int count, A first...) { va_list ap; va_start(ap, first); A x = va_arg(ap, A); } int main() { A a, b, c; f(3, a, b, c); } g++ -v -W -Wall -save-temps a.C Reading specs from /usr/local/lib/gcc/i686-pc-linux-gnu/3.4/specs Configured with: ../gcc/configure --enable-threads=posix --enable-languages=c,c++,f77,objc --enable-__cxa_atexit --enable-debug Thread model: posix gcc version 3.4 20030806 (experimental) /usr/local/libexec/gcc/i686-pc-linux-gnu/3.4/cc1plus -E -quiet -v -D_GNU_SOURCE a.C -mtune=pentiumpro -W -Wall -o a.ii ignoring nonexistent directory "NONE/include" ignoring nonexistent directory "/usr/local/lib/gcc/i686-pc-linux-gnu/3.4/../../../../i686-pc-linux-gnu/include" #include "..." search starts here: #include <...> search starts here: /usr/local/lib/gcc/i686-pc-linux-gnu/3.4/../../../../include/c++/3.4 /usr/local/lib/gcc/i686-pc-linux-gnu/3.4/../../../../include/c++/3.4/i686-pc-linux-gnu /usr/local/lib/gcc/i686-pc-linux-gnu/3.4/../../../../include/c++/3.4/backward /usr/local/include /usr/local/lib/gcc/i686-pc-linux-gnu/3.4/include /usr/include End of search list. /usr/local/libexec/gcc/i686-pc-linux-gnu/3.4/cc1plus -fpreprocessed a.ii -quiet -dumpbase a.C -mtune=pentiumpro -auxbase a -W -Wall -version -o a.s GNU C++ version 3.4 20030806 (experimental) (i686-pc-linux-gnu) compiled by GNU C version 3.4 20030806 (experimental). GGC heuristics: --param ggc-min-expand=47 --param ggc-min-heapsize=31899 a.C: In function `void f(int, A, ...)': a.C:14: warning: cannot pass objects of non-POD type `class A' through `...'; call will abort at runtime a.C:15: warning: cannot receive objects of non-POD type `class A' through `...' a.C:14: warning: second parameter of `va_start' not last named argument a.C:15: internal compiler error: in cp_expr_size, at cp/cp-lang.c:330 Please submit a full bug report, with preprocessed source if appropriate. See <URL:http://gcc.gnu.org/bugs.html> for instructions. # 1 "a.C" # 1 "<built-in>" # 1 "<command line>" # 1 "a.C" # 1 "/usr/local/lib/gcc/i686-pc-linux-gnu/3.4/include/stdarg.h" 1 3 4 # 43 "/usr/local/lib/gcc/i686-pc-linux-gnu/3.4/include/stdarg.h" 3 4 typedef __builtin_va_list __gnuc_va_list; # 105 "/usr/local/lib/gcc/i686-pc-linux-gnu/3.4/include/stdarg.h" 3 4 typedef __gnuc_va_list va_list; # 2 "a.C" 2 class A { public: A(); A(const A& x); A& operator= (const A& a); ~A(); }; void f(int count, A first...) { va_list ap; __builtin_va_start(ap,first); A x = __builtin_va_arg(ap,A); } int main() { A a, b, c; f(3, a, b, c); }
The ICE is certainly a regression. I guess the warning is warranted, since you really put objects of non-POD type through ... W.
From Phil's regression hunter: Search converges between 2003-01-02-trunk (#182) and 2003- 01-03-trunk (#183).
The regression in PR 11971 was introduced or exposed by this patch: --- gcc/gcc/ChangeLog --- 2003-01-02 Jason Merrill <jason@redhat.com> * fold-const.c (fold) [COND_EXPR]: Avoid NOP_EXPRs better. * integrate.c (copy_decl_for_inlining): Don't clear the rtl for static/external decls. --- gcc/gcc/cp/ChangeLog --- 2003-01-02 Jason Merrill <jason@redhat.com> * call.c (build_conditional_expr): Stabilize lvalues properly. * cvt.c (ocp_convert): Don't build NOP_EXPRs of class type. * tree.c (lvalue_p_1): Don't allow sloppy NOP_EXPRs as lvalues. Don't allow CALL_EXPR or VA_ARG_EXPR, either. * call.c (convert_like_real): Call decl_constant_value for an IDENTITY_CONV even if there are no more conversions. [also cvt.c, decl2.c, dump.c, except.c, init.c, tree.c] The regression hunt took place on i686-pc-linux-gnu using the submitter's test case. Before this patch the test case got these additional warnings rather than the ICE: 11971.C: In function `int main()': 11971.C:22: warning: cannot pass objects of non-POD type `class A' through `...'; call will abort at runtime 11971.C:22: warning: cannot pass objects of non-POD type `class A' through `...'; call will abort at runtime
Here is a valid program which we currently reject because va_start isn't magic enough. We currently treat it as a normal variadic function. #include <stdarg.h> class A { public: A() {} A(const A& x) {} A& operator= (const A& a) {} ~A() {} }; void f(A first...) { va_list ap; va_start(ap, first); int x = va_arg(ap, int); } int main() { A a; f(a, 3); }
*** Bug 11929 has been marked as a duplicate of this bug. ***
Subject: Bug 11971 CVSROOT: /cvs/gcc Module name: gcc Changes by: jason@gcc.gnu.org 2003-12-09 04:08:50 Modified files: gcc/cp : ChangeLog call.c cp-tree.h ptree.c tree.c Log message: PR c++/11971 * tree.c (build_local_temp): Split out from build_cplus_new. (force_target_expr): New fn. * call.c (call_builtin_trap): Call it. Take a type parm. (convert_arg_to_ellipsis): Pass it. (build_x_va_arg): Use call_builtin_trap. PR c++/11929 * call.c (magic_varargs_p): New fn. (build_over_call): Do no ellipsis conversions for arguments to functions with magic varargs. Patches: http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/ChangeLog.diff?cvsroot=gcc&r1=1.3789&r2=1.3790 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/call.c.diff?cvsroot=gcc&r1=1.442&r2=1.443 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/cp-tree.h.diff?cvsroot=gcc&r1=1.936&r2=1.937 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/ptree.c.diff?cvsroot=gcc&r1=1.37&r2=1.38 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/tree.c.diff?cvsroot=gcc&r1=1.355&r2=1.356
Fixed for 3.4.