Sender: rich AT phekda DOT freeserve DOT co DOT uk Message-ID: <3CF106B5.55044328@phekda.freeserve.co.uk> Date: Sun, 26 May 2002 17:00:53 +0100 From: Richard Dawe X-Mailer: Mozilla 4.77 [en] (X11; U; Linux 2.2.19 i586) X-Accept-Language: de,fr MIME-Version: 1.0 To: DJGPP workers Subject: gcc 3.1 and some test suite problems Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Reply-To: djgpp-workers AT delorie DOT com Hello. I've just been looking at building the test suite with gcc 3.1. Most of the problems I found were to do with headers not being included (e.g. for strlen()). But the following have no obvious fix to me: tests/libc/ansi/stdarg/stdarg.c: This needs __dj_va_rounded_size, but it's not found. Presumably gcc 3.1 uses its own va_*, so we never pull in __dj_va_rounded_size. Should the test be amended to only use __dj_va_rounded_size if the gcc version is less than 3? tests/libc/dos/ffxx.cc: This won't link, presumably because some C++ library hasn't been linked in. I tried including libstdcxx.a and libsupcxx.a on the link line, but that didn't seem to help. tests/libc/go32/signals.c: The following line: urand = ((double)rand()) / RAND_MAX; gives this warning: gcc @../../gcc.opt -I. -I- -I../../../include -pg -c signals.c cc1.exe: warnings being treated as errors signals.c: In function `int_handler': signals.c:23: warning: cast does not match function type I don't understand this. Casting to a long, then a double seems to work. (I don't have K&R to hand, to help me out.) tests/libc/go32/makefile: This adds '-pg' to LDFLAGS, which ld from GNU binutils 2.11.2 chokes on. tests/libc/pc_hw/farptr/farptr.c: main() does not have argc, argv as usual, but a weird triplet of values. This looks intentional, but how do you run it? Bye, Rich =] -- Richard Dawe [ http://www.phekda.freeserve.co.uk/richdawe/ ]