On Thu, 2024-12-19 at 12:17 -0600, Eric Blake wrote: > Currently GNU make does not conform to the required .c.o inference > rule because it adds a -o option.
Technically, make's built-in recipe for .c.o is: $(COMPILE.c) $(OUTPUT_OPTION) $< where OUTPUT_OPTION is set to "-o $@" if GNU Make's configure-time check determined that cc supports the -o option, and empty otherwise. This has always been hinky because obviously just because the cc that configure discovered at build time supports (or not) -o, doesn't mean that whatever C compiler you end up using in your makefile does too (or doesn't either). But, it works (probably because there hasn't been a C compiler that _didn't_ handle -o for decades). Note, your example is wrong because the POSIXLY_CORRECT environment variable doesn't control whether make is supposed to run in POSIX mode or not (that would be completely unworkable, if the setting/unsetting an environment variable changed the way make interpreted the current makefile!) You have to set the .POSIX: special target in your makefile if you want it to have POSIX syntax; something like: $ touch foo.c $ echo '.POSIX:' | make -f- -n foo.o Unfortunately this gives the same result. There is actually code in GNU Make to reset OUTPUT_OPTION to empty if POSIX is specified... but it's commented out. I apparently added it, and commented it out, back in 2006 and the explanation 2006 me gave makes no sense to 2024 me, so one of us is dumb :). In the abstract I would not have a problem with uncommenting this code, assuming I can come to grips with why I commented it out before. > What's worse, the addition of a -o rule causes interesting effects; > $< contains a name with a /, the presence or absence of .o changes > which directory the compiler would put its outpt in, which in turn > influences where make should look for any existing .o file to check > for timestamps before deciding if compilation is needed. I don't understand this, and I didn't really understand the part of the but report either. If someone could provide more details about exactly what the problem is I'd be interested to learn. Thanks Eric!
