From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 20507 invoked by alias); 23 Mar 2002 13:26:02 -0000 Mailing-List: contact gcc-prs-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Archive: List-Post: List-Help: Sender: gcc-prs-owner@gcc.gnu.org Received: (qmail 20484 invoked by uid 71); 23 Mar 2002 13:26:01 -0000 Date: Sat, 23 Mar 2002 05:26:00 -0000 Message-ID: <20020323132601.20483.qmail@sources.redhat.com> To: neil@gcc.gnu.org Cc: gcc-prs@gcc.gnu.org, From: Neil Booth Subject: Re: preprocessor/3951: need for -MF w/ -M breaks backward compat. w/ documented features. Reply-To: Neil Booth X-SW-Source: 2002-03/txt/msg00874.txt.bz2 List-Id: The following reply was made to PR preprocessor/3951; it has been noted by GNATS. From: Neil Booth To: cgd@broadcom.com Cc: gcc-gnats@gcc.gnu.org Subject: Re: preprocessor/3951: need for -MF w/ -M breaks backward compat. w/ documented features. Date: Sat, 23 Mar 2002 13:16:56 +0000 I've cc-ed GNATS. Chris G. Demetriou wrote:- > On further testing, not sure that all is happy in 3.0.x with the > matches that you made. (I didn't cc the list/gnats on this, but if > you'd like feel free to fwd this msg there.) > > > It could be that i've goofed them up applying them to my source tree, > but I don't _think_ I did. > > > For a file foo.c and foo.S with 'junk' contents, what does: > > gcc -MMD -c -o foo.o $@ > > produce for you? > > similarly, what if you say "-E" instead of -c? > > > in my little world, for the .c / -c case, things are fine. > > for the .c / -E, .S / -c, and .S / -E cases, the the preprocessed > output is zero-length (leading to ... not the right thing in the .S / > -c case, especially 8-). Wah! It works now! And with -save-temps!! Happy!!! I've done everything I can think of to break it, and compared everything with 2.95.4; and it's all matching. The final patch I sent you seems to be the holy grail. It gets the above right, fixes the 2.95 issues we wanted to avoid, and handles -MF, -MQ and -MT like we want. For the benefit of GNATS, I'll reproduce the patch below. If you could confirm this fixes this issue once and for all, I'll apply it to all 3 trees. Thanks for persevering! Neil. Index: gcc.c =================================================================== RCS file: /cvs/gcc/gcc/gcc/gcc.c,v retrieving revision 1.205.2.26 diff -u -p -r1.205.2.26 gcc.c --- gcc.c 2002/03/11 21:38:14 1.205.2.26 +++ gcc.c 2002/03/23 12:07:37 @@ -596,8 +596,8 @@ static const char *cpp_options = "%{C:%{!E:%eGNU C does not support -C without using -E}}\ %{std*} %{nostdinc*}\ %{C} %{v} %{I*} %{P} %{$} %I\ - %{MD:-M -MF %W{!o: %b.d}%W{o*:%.d%*}}\ - %{MMD:-MM -MF %W{!o: %b.d}%W{o*:%.d%*}}\ + %{MD:-MD %W{!o: %b.d}%W{o*:%.d%*}}\ + %{MMD:-MMD %W{!o: %b.d}%W{o*:%.d%*}}\ %{M} %{MM} %W{MF*} %{MG} %{MP} %{MQ*} %{MT*}\ %{!E:%{!M:%{!MM:%{MD|MMD:%{o*:-MQ %*}}}}}\ %{!no-gcc:-D__GNUC__=%v1 -D__GNUC_MINOR__=%v2 -D__GNUC_PATCHLEVEL__=%v3}\ Index: cppinit.c =================================================================== RCS file: /cvs/gcc/gcc/gcc/cppinit.c,v retrieving revision 1.147.2.12 diff -u -p -r1.147.2.12 cppinit.c --- cppinit.c 2002/03/11 21:38:18 1.147.2.12 +++ cppinit.c 2002/03/23 12:07:41 @@ -1459,10 +1459,16 @@ cpp_handle_option (pfile, argc, argv) CPP_OPTION (pfile, print_deps_missing_files) = 1; break; case OPT_M: + /* When doing dependencies with -M or -MM, suppress normal + preprocessed output, but still do -dM etc. as software + depends on this. Preprocessed output occurs if -MD, -MMD + or environment var dependency generation is used. */ CPP_OPTION (pfile, print_deps) = 2; + CPP_OPTION (pfile, no_output) = 1; break; case OPT_MM: CPP_OPTION (pfile, print_deps) = 1; + CPP_OPTION (pfile, no_output) = 1; break; case OPT_MF: CPP_OPTION (pfile, deps_file) = arg; @@ -1476,12 +1482,6 @@ cpp_handle_option (pfile, argc, argv) deps_add_target (pfile->deps, arg, opt_code == OPT_MQ); break; - /* -MD and -MMD for cpp0 are deprecated and undocumented - (use -M or -MM with -MF instead), and probably should be - removed with the next major GCC version. For the moment - we allow these for the benefit of Automake 1.4, which - uses these when dependency tracking is enabled. Automake - 1.5 will fix this. */ case OPT_MD: CPP_OPTION (pfile, print_deps) = 2; CPP_OPTION (pfile, deps_file) = arg; @@ -1786,10 +1786,6 @@ init_dependency_output (pfile) /* If -M or -MM was seen, default output to wherever was specified with -o. out_fname is non-NULL here. */ CPP_OPTION (pfile, deps_file) = CPP_OPTION (pfile, out_fname); - - /* When doing dependencies, suppress normal preprocessed output. - Still do -dM, -dI etc. as e.g. glibc depends on this. */ - CPP_OPTION (pfile, no_output) = 1; } static void