Message ID | 20151008215558.C91542C3AA0@topped-with-meat.com |
---|---|
State | New |
Headers | show |
On Thu, 8 Oct 2015, Roland McGrath wrote: > > It seems like that is still a bug in the Makefile though, should I > > submit a patch? > > I just committed this myself: > > 2015-10-08 Roland McGrath <roland@hack.frob.com> > > [BZ #18872] > * stdio-common/Makefile ($(objpfx)tst-unbputc.out): > Move out from under [$(run-built-tests) = yes] conditional. That looks wrong. tst-unbputc.sh uses test_program_prefix to run a newly built test.
> > 2015-10-08 Roland McGrath <roland@hack.frob.com> > > > > [BZ #18872] > > * stdio-common/Makefile ($(objpfx)tst-unbputc.out): > > Move out from under [$(run-built-tests) = yes] conditional. > > That looks wrong. tst-unbputc.sh uses test_program_prefix to run a newly > built test. Oh, it was unrelated. I failed to notice that. But why is it wrong? I don't see why we need to avoid defining the targets. Adding them to tests-special is conditionalized, so nothing will depend on that.
On Thu, 8 Oct 2015, Roland McGrath wrote: > > > 2015-10-08 Roland McGrath <roland@hack.frob.com> > > > > > > [BZ #18872] > > > * stdio-common/Makefile ($(objpfx)tst-unbputc.out): > > > Move out from under [$(run-built-tests) = yes] conditional. > > > > That looks wrong. tst-unbputc.sh uses test_program_prefix to run a newly > > built test. > > Oh, it was unrelated. I failed to notice that. But why is it wrong? I > don't see why we need to avoid defining the targets. Adding them to > tests-special is conditionalized, so nothing will depend on that. In that case it should be OK (but maybe we should be consistent, in whatever direction, about when such rules are / aren't conditionalized).
> In that case it should be OK (but maybe we should be consistent, in > whatever direction, about when such rules are / aren't conditionalized). I think consistency is good. I think we should favor not conditionalizing anywhere that we do not have to, just to reduce clutter in the makefiles.
On 10/08/2015 06:24 PM, Roland McGrath wrote: >> In that case it should be OK (but maybe we should be consistent, in >> whatever direction, about when such rules are / aren't conditionalized). > > I think consistency is good. I think we should favor not conditionalizing > anywhere that we do not have to, just to reduce clutter in the makefiles. +1. c.
diff --git a/stdio-common/Makefile b/stdio-common/Makefile index 40f5cdf..df4122d 100644 --- a/stdio-common/Makefile +++ b/stdio-common/Makefile @@ -74,7 +74,6 @@ include ../Rules tst-printf-bz18872-ENV = MALLOC_TRACE=$(objpfx)tst-printf-bz18872.mtrace -ifeq ($(run-built-tests),yes) $(objpfx)tst-unbputc.out: tst-unbputc.sh $(objpfx)tst-unbputc $(SHELL) $< $(common-objpfx) '$(test-program-prefix)'; \ $(evaluate-test) @@ -91,7 +90,6 @@ $(objpfx)tst-printf-bz18872.c: tst-printf-bz18872.sh $(objpfx)tst-printf-bz18872-mem.out: $(objpfx)tst-printf-bz18872.out $(common-objpfx)malloc/mtrace $(objpfx)tst-printf-bz18872.mtrace > $@; \ $(evaluate-test) -endif CFLAGS-vfprintf.c = -Wno-uninitialized CFLAGS-vfwprintf.c = -Wno-uninitialized