Re: [PATCH v2] Strip build path directories in tools and hypervisor

[Marek Marczykowski-Górecki <marmarek@xxxxxxxxxxxxxxxxxxxxxx>]

On Tue, Sep 22, 2026 at 03:06:35PM +0200, Jan Beulich wrote:
> On 04.09.2025 13:41, Marek Marczykowski-Górecki wrote:
> > Use -fdebug-prefix-map in preference to -ffile-prefix-map, as it's
> > available in earlier toolchain versions. But use it together with
> > -fmacro-prefix-map (if available) for hypervisor build, otherwise it
> > still contains some paths in out-of-tree builds.
> > 
> > The out of tree build requires -fdebug-prefix-map mapping for both source
> > dir and object dir - otherwise the latter is included (2 occurrences) in
> > xen-syms. Note the ./xen path for out of tree builds may not be strictly
> > correct choice, but it's consistent across the tree, and just require
> > starting debugger from the source, not object, directory.
> > 
> > Ensure to have a realpath for XEN_ROOT else it fails to substitute
> > properly paths in strings sections.
> > 
> > Signed-off-by: Marek Marczykowski-Górecki <marmarek@xxxxxxxxxxxxxxxxxxxxxx>
> 
> Coming back to this, as it was also mentioned at the summit.
> 
> > --- a/tools/Makefile
> > +++ b/tools/Makefile
> > @@ -1,4 +1,4 @@
> > -XEN_ROOT = $(CURDIR)/..
> > +XEN_ROOT = $(realpath $(CURDIR)/..)
> 
> $(realpath ...) may not be available (with older make), it wants to be
> $(call realpath, ...), allowing use of the fallback in ./Config.mk.
> 
> I'm further not really sure if $(realpath ...) is actually meant here;
> to me $(abspath ...) would seem more logical (as we don't want to
> resolve symlinks in the path, but rather keep it as specified / seen).

Honestly, I don't remember all the details a year later... But I think,
based on the commit message, there was some case where XEN_ROOT needed
to textually match the actual directory, possibly when some other
part of the build system embedded a resolved path to the binary anyway.

> And finally - isn't the above rendering redundant ...
> 
> > --- a/tools/Rules.mk
> > +++ b/tools/Rules.mk
> > @@ -166,6 +166,8 @@ endif
> >  CFLAGS-$(CONFIG_X86_32) += $(call cc-option,$(CC),-mno-tls-direct-seg-refs)
> >  CFLAGS += $(CFLAGS-y)
> >  
> > +$(call cc-option-add,CFLAGS,CC,-fdebug-prefix-map=$(realpath 
> > $(XEN_ROOT))=.)
> 
> ... the use of $(realpath ...) here?

This seems to be correct observation.

-- 
Best Regards,
Marek Marczykowski-Górecki
Invisible Things Lab

Attachment: signature.asc
Description: PGP signature