Hi, I'm trying to get the llvm-test suite going on Fedora Core 5 and I have a number of failures with the current HEAD. The major one (see PR879) has a workaround. Now, I'm getting a conflict in the CBakcned while compiling siod (and others): gcc Output/siod.cbe.c -lm -ldl -fno-strict-aliasing -O2 -o Output/siod.cbe Output/siod.cbe.c:471: error: conflicting types for '_setjmp' /usr/include/setjmp.h:63: error: previous declaration of '_setjmp' was here siod.cbe.c:471: int _setjmp(struct l_struct_2E___jmp_buf_tag *); /usr/include/setjmp.h:64: extern int _setjmp (struct __jmp_buf_tag __env[1]) __THROW; I assume this is because of the array vs. the pointer. Why does the C Backend provide its own definition of _setjmp when it knows there is one in the header files vi the HAVE__SETJMP #define? Reid.
On Fri, 25 Aug 2006, Reid Spencer wrote:> I'm trying to get the llvm-test suite going on Fedora Core 5 and I have > a number of failures with the current HEAD. The major one (see PR879) > has a workaround. Now, I'm getting a conflict in the CBakcned while > compiling siod (and others): > I assume this is because of the array vs. the pointer. Why does the C > Backend provide its own definition of _setjmp when it knows there is one > in the header files vi the HAVE__SETJMP #define?I don't know, but this may be related to rev 1.265 of CBackend/Writer.cpp. -Chris -- http://nondot.org/sabre/ http://llvm.org/