Re: [PATCH] kbuild: Allow a suffix with $(LLVM)

[Nathan Chancellor]

On Fri, Mar 04, 2022 at 08:16:00PM +0900, Masahiro Yamada wrote:
> On Thu, Mar 3, 2022 at 8:47 AM Nathan Chancellor <nathan@xxxxxxxxxx> wrote:
> >
> > The LLVM make variable allows a developer to quickly switch between the
> > GNU and LLVM tools. However, it does not handle versioned binaries, such
> > as the ones shipped by Debian, as LLVM=1 just defines the tool variables
> > with the unversioned binaries.
> >
> > There was some discussion during the review of the patch that introduces
> > LLVM=1 around versioned binaries, ultimately coming to the conclusion
> > that developers can just add the folder that contains the unversioned
> > binaries to their PATH, as Debian's versioned suffixed binaries are
> > really just symlinks to the unversioned binaries in /usr/lib/llvm-#/bin:
> >
> > $ realpath /usr/bin/clang-14
> > /usr/lib/llvm-14/bin/clang
> >
> > $ PATH=/usr/lib/llvm-14/bin:$PATH make ... LLVM=1
> >
> > However, that can be cumbersome to developers who are constantly testing
> > series with different toolchains and versions. It is simple enough to
> > support these versioned binaries directly in the Kbuild system by
> > allowing the developer to specify the version suffix with LLVM=, which
> > is shorter than the above suggestion:
> >
> > $ make ... LLVM=-14
> >
> > It does not change the meaning of LLVM=1 (which will continue to use
> > unversioned binaries) and it does not add too much additional complexity
> > to the existing $(LLVM) code, while allowing developers to quickly test
> > their series with different versions of the whole LLVM suite of tools.
> >
> > Link: https://lore.kernel.org/r/20200317215515.226917-1-ndesaulniers@xxxxxxxxxx/
> > Link: https://lore.kernel.org/r/20220224151322.072632223@xxxxxxxxxxxxx/
> > Suggested-by: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
> > Reviewed-by: Kees Cook <keescook@xxxxxxxxxxxx>
> > Reviewed-by: Nick Desaulniers <ndesaulniers@xxxxxxxxxx>
> > Signed-off-by: Nathan Chancellor <nathan@xxxxxxxxxx>
> > ---
> >
> > RFC -> v1: https://lore.kernel.org/r/Yh%2FegU1LZudfrgVy@dev-arch.thelio-3990X/
> >
> > * Tidy up commit message slightly.
> >
> > * Add tags.
> >
> > * Add links to prior discussions for context.
> >
> > * Add change to tools/testing/selftests/lib.mk.
> >
> > I would like for this to go through the Kbuild tree, please ack as
> > necessary.
> >
> >  Documentation/kbuild/llvm.rst  |  7 +++++++
> >  Makefile                       | 24 ++++++++++++++----------
> >  tools/scripts/Makefile.include | 20 ++++++++++++--------
> >  tools/testing/selftests/lib.mk |  6 +++++-
> >  4 files changed, 38 insertions(+), 19 deletions(-)
> >
> > diff --git a/Documentation/kbuild/llvm.rst b/Documentation/kbuild/llvm.rst
> > index d32616891dcf..5805a8473a36 100644
> > --- a/Documentation/kbuild/llvm.rst
> > +++ b/Documentation/kbuild/llvm.rst
> > @@ -60,6 +60,13 @@ They can be enabled individually. The full list of the parameters: ::
> >           OBJCOPY=llvm-objcopy OBJDUMP=llvm-objdump READELF=llvm-readelf \
> >           HOSTCC=clang HOSTCXX=clang++ HOSTAR=llvm-ar HOSTLD=ld.lld
> >
> > +If your LLVM tools have a suffix and you prefer to test an explicit version rather
> > +than the unsuffixed executables, use ``LLVM=<suffix>``. For example: ::
> > +
> > +       make LLVM=-14
> > +
> > +will use ``clang-14``, ``ld.lld-14``, etc.
> > +
> >  The integrated assembler is enabled by default. You can pass ``LLVM_IAS=0`` to
> >  disable it.
> 
> 
> Perhaps, it might be worth mentioning the difference between
> LLVM=1 and LLVM=<suffix>
> 
> The current behavior is,
> any value other than '1' is regarded as a suffix.

Maybe just adding something like:

"... prefer to test an explicit version rather than the unsuffixed
executables like above, ..."

? I'll try to come up with a clearer way to word everything for v2.

> > diff --git a/Makefile b/Makefile
> > index a82095c69fdd..963840c00eae 100644
> > --- a/Makefile
> > +++ b/Makefile
> > @@ -424,8 +424,12 @@ HOST_LFS_LDFLAGS := $(shell getconf LFS_LDFLAGS 2>/dev/null)
> >  HOST_LFS_LIBS := $(shell getconf LFS_LIBS 2>/dev/null)
> >
> >  ifneq ($(LLVM),)
> > -HOSTCC = clang
> > -HOSTCXX        = clang++
> > +ifneq ($(LLVM),1)
> > +LLVM_SFX := $(LLVM)
> > +endif
> 
> I am OK with this, but please note LLVM=0
> uses 'clang0', 'ld.lld0' instead of disabling
> LLVM explicitly.
> 
> This might be a small surprise because LLVM_IAS=0
> is used to disable the integrated assembler.

Right, but we have that problem right now, as LLVM=0 is currently
treated like LLVM=1. I suppose I could add a line to the documentation
in a follow up change to clarify this.

> If you want handle LLVM=<suffix>
> only when <suffix> start with a hyphen,
> you can do like this:
> 
> ifneq ($(filter -%, $(LLVM)),)
> LLVM_SFX := $(LLVM)
> endif

I did think about this. I guess the only reason I did not do that in
this version is someone might have a different suffix scheme than
Debian's but it is probably better to be a little bit more precise based
on what we know in this moment. I will change it to that and update the
documentation.

> In the future, If somebody requests to support
>     make LLVM=/path/to/my/own/llvm/dir/
> to use llvm tools in that path,
> we can expand the code like this:
> 
> 
> 
> # "LLVM=foo/bar/" is a syntax sugar of "LLVM=1 LLVM_PFX=foo/bar"
> # "LLVM=-foo" is a syntax sugar of "LLVM=1 LLVM_SFX=-foo"
> 
> ifneq ($(filter %/, $(LLVM)),)
> LLVM_PFX := $(LLVM)
> else ifneq ($(filter -%, $(LLVM)),)
> LLVM_SFX := $(LLVM)
> endif

I know I personally I would use the prefix form when testing with LLVM=1
so I think I will just go ahead and support that now, especially since
Peter had aimed to support a full path with his CC patch that we NAK'd.

> Lastly, I personally prefer to fully spell LLVM_SUFFIX
> as Nick originally suggested:
> https://lkml.org/lkml/2020/3/17/1477

Ack, I have changed this locally and I'll send a v2 along shortly once I
have written some documentation to codify these suggested changes.

Thank you for the comments and review, cheers!
Nathan