Re: [PATCH v1 01/11] perf build: Rename NO_DWARF to NO_LIBDW

[Namhyung Kim]

On Thu, Sep 26, 2024 at 08:02:45AM -0700, Ian Rogers wrote:
> On Wed, Sep 25, 2024 at 5:26 PM Namhyung Kim <namhyung@xxxxxxxxxx> wrote:
> >
> > On Tue, Sep 24, 2024 at 09:04:08AM -0700, Ian Rogers wrote:
> > > NO_DWARF could mean more than NO_LIBDW support, in particular no
> > > libunwind support. Rename to be more intention revealing.
> > >
> > > Signed-off-by: Ian Rogers <irogers@xxxxxxxxxx>
> > > ---
> > >  tools/perf/Makefile.config         | 16 ++++++++--------
> > >  tools/perf/Makefile.perf           |  2 +-
> > >  tools/perf/arch/arm/Makefile       |  2 +-
> > >  tools/perf/arch/arm64/Makefile     |  2 +-
> > >  tools/perf/arch/csky/Makefile      |  2 +-
> > >  tools/perf/arch/loongarch/Makefile |  2 +-
> > >  tools/perf/arch/mips/Makefile      |  2 +-
> > >  tools/perf/arch/powerpc/Makefile   |  2 +-
> > >  tools/perf/arch/riscv/Makefile     |  2 +-
> > >  tools/perf/arch/s390/Makefile      |  2 +-
> > >  tools/perf/arch/sh/Makefile        |  2 +-
> > >  tools/perf/arch/sparc/Makefile     |  2 +-
> > >  tools/perf/arch/x86/Makefile       |  2 +-
> > >  tools/perf/arch/xtensa/Makefile    |  2 +-
> > >  tools/perf/builtin-probe.c         |  2 +-
> > >  15 files changed, 22 insertions(+), 22 deletions(-)
> > >
> > > diff --git a/tools/perf/Makefile.config b/tools/perf/Makefile.config
> > > index 5e26d3a91b36..55a39211496d 100644
> > > --- a/tools/perf/Makefile.config
> > > +++ b/tools/perf/Makefile.config
> > > @@ -429,7 +429,7 @@ ifeq ($(feature-file-handle), 1)
> > >  endif
> > >
> > >  ifdef NO_LIBELF
> > > -  NO_DWARF := 1
> > > +  NO_LIBDW := 1
> > >    NO_LIBUNWIND := 1
> > >    NO_LIBDW_DWARF_UNWIND := 1
> > >    NO_LIBBPF := 1
> > > @@ -471,9 +471,9 @@ else
> > >        endif
> > >      endif
> > >      ifneq ($(feature-dwarf), 1)
> > > -      ifndef NO_DWARF
> > > +      ifndef NO_LIBDW
> > >          $(warning No libdw.h found or old libdw.h found or elfutils is older than 0.138, disables dwarf support. Please install new elfutils-devel/libdw-dev)
> > > -        NO_DWARF := 1
> > > +        NO_LIBDW := 1
> > >        endif
> > >      else
> > >        ifneq ($(feature-dwarf_getlocations), 1)
> > > @@ -496,7 +496,7 @@ ifeq ($(feature-libaio), 1)
> > >    endif
> > >  endif
> > >
> > > -ifdef NO_DWARF
> > > +ifdef NO_LIBDW
> > >    NO_LIBDW_DWARF_UNWIND := 1
> > >  endif
> > >
> > > @@ -574,17 +574,17 @@ ifndef NO_LIBELF
> > >      endif
> > >    endif
> > >
> > > -  ifndef NO_DWARF
> > > +  ifndef NO_LIBDW
> > >      ifeq ($(origin PERF_HAVE_DWARF_REGS), undefined)
> > >        $(warning DWARF register mappings have not been defined for architecture $(SRCARCH), DWARF support disabled)
> > > -      NO_DWARF := 1
> > > +      NO_LIBDW := 1
> > >      else
> > >        CFLAGS += -DHAVE_DWARF_SUPPORT $(LIBDW_CFLAGS)
> > >        LDFLAGS += $(LIBDW_LDFLAGS)
> > >        EXTLIBS += ${DWARFLIBS}
> > >        $(call detected,CONFIG_DWARF)
> > >      endif # PERF_HAVE_DWARF_REGS
> > > -  endif # NO_DWARF
> > > +  endif # NO_LIBDW
> > >
> > >    ifndef NO_LIBBPF
> > >      ifeq ($(feature-bpf), 1)
> > > @@ -633,7 +633,7 @@ ifdef PERF_HAVE_JITDUMP
> > >  endif
> > >
> > >  ifeq ($(SRCARCH),powerpc)
> > > -  ifndef NO_DWARF
> > > +  ifndef NO_LIBDW
> > >      CFLAGS += -DHAVE_SKIP_CALLCHAIN_IDX
> > >    endif
> > >  endif
> > > diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefile.perf
> > > index 9dd2e8d3f3c9..a144bfaf8aeb 100644
> > > --- a/tools/perf/Makefile.perf
> > > +++ b/tools/perf/Makefile.perf
> > > @@ -40,7 +40,7 @@ include ../scripts/utilities.mak
> > >  #
> > >  # Define EXTRA_PERFLIBS to pass extra libraries to PERFLIBS.
> > >  #
> > > -# Define NO_DWARF if you do not want debug-info analysis feature at all.
> > > +# Define NO_LIBDW if you do not want debug-info analysis feature at all.
> >
> > Can we keep NO_DWARF for compatibility and set NO_LIBDW=1 internally?
> > I think it's fine to change it here to advertise NO_LIBDW over NO_DWARF
> > but still want to support NO_DWARF as well.
> 
> This would be actively confusing. NO_DWARF should really imply
> NO_LIBUNWIND, it is like removing the ability for "perf record
> --call-graph dwarf" to work - dwarf meaning libdw or libunwind in the
> tool. Having NO_DWARF control NO_LIBDW, I'm not sure what I'm being
> asked to do. The point of this clean up is to make the names mean what
> they should mean and not something potentially confusingly broader,
> but you're asking for the confusingly broader case to mean just libdw.
> I don't follow and I don't see how it serves the purpose of the
> cleanup.

I have no idea how many people use NO_DWARF actually.  We can probably
ignore such users and convert it to NO_LIBDW.

What I actually thought was

  ifeq ($(NO_DWARF),1)
    NO_LIBDW := 1
    NO_LIBUNWIND := 1
  endif

but it'd break the compatibility too. :(

Thanks,
Namhyung