| Message ID | 75d94bf1-b419-8a82-2d15-fb02e56109d8@suse.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | xen: beginnings of moving library-like code into an archive | expand |
On Mon, Sep 14, 2020 at 12:15:39PM +0200, Jan Beulich wrote: > Switch to $(call if_changed,ld) where possible; presumably not doing so > in e321576f4047 ("xen/build: start using if_changed") right away was an > oversight, as it did for Arm in (just) one case. It failed to add > prelink.o to $(targets), though, causing - judging from the observed > behavior on x86 - undue rebuilds of the final binary (because of > prelink.o getting rebuild for $(cmd_prelink.o) being empty, in turn > because of .prelink.o.cmd not getting read) during "make install-xen". I'm not sure I follow why prelink.o needs to be added to targets, does this offer some kind of protection against rebuilds when doing make install? The switch to if_changed LGTM. Thanks, Roger.
On 15.09.2020 13:56, Roger Pau Monné wrote: > On Mon, Sep 14, 2020 at 12:15:39PM +0200, Jan Beulich wrote: >> Switch to $(call if_changed,ld) where possible; presumably not doing so >> in e321576f4047 ("xen/build: start using if_changed") right away was an >> oversight, as it did for Arm in (just) one case. It failed to add >> prelink.o to $(targets), though, causing - judging from the observed >> behavior on x86 - undue rebuilds of the final binary (because of >> prelink.o getting rebuild for $(cmd_prelink.o) being empty, in turn >> because of .prelink.o.cmd not getting read) during "make install-xen". > > I'm not sure I follow why prelink.o needs to be added to targets, does > this offer some kind of protection against rebuilds when doing make > install? In a way, but (as I view it) not really. It is the use of ... > The switch to if_changed LGTM. ... if_changed which requires this. .*.cmd files will only be loaded for anything explicitly or implicitly listed as a target. While .o coming from $(obj-y) get added there automatically, prelink.o is not something that could be recognized as needing adding, hence the "manual" insertion. Without .prelink.o.cmd loaded, $(if_changed ) will always arrange for it to get re-built, because it then will consider the command used to build the file to have changed (as the stored one appears to be empty). Jan
On Tue, Sep 15, 2020 at 02:26:34PM +0200, Jan Beulich wrote: > On 15.09.2020 13:56, Roger Pau Monné wrote: > > On Mon, Sep 14, 2020 at 12:15:39PM +0200, Jan Beulich wrote: > >> Switch to $(call if_changed,ld) where possible; presumably not doing so > >> in e321576f4047 ("xen/build: start using if_changed") right away was an > >> oversight, as it did for Arm in (just) one case. It failed to add > >> prelink.o to $(targets), though, causing - judging from the observed > >> behavior on x86 - undue rebuilds of the final binary (because of > >> prelink.o getting rebuild for $(cmd_prelink.o) being empty, in turn > >> because of .prelink.o.cmd not getting read) during "make install-xen". > > > > I'm not sure I follow why prelink.o needs to be added to targets, does > > this offer some kind of protection against rebuilds when doing make > > install? > > In a way, but (as I view it) not really. It is the use of ... > > > The switch to if_changed LGTM. > > ... if_changed which requires this. .*.cmd files will only be loaded > for anything explicitly or implicitly listed as a target. While .o > coming from $(obj-y) get added there automatically, prelink.o is not > something that could be recognized as needing adding, hence the > "manual" insertion. This seems very prone to mistakes, as you have to remember that whatever object that uses if_changed should also be part of the targets set, or else it will get rebuild unconditionally. I think adding some of the above reasoning to the commit message would be helpful IMO. > Without .prelink.o.cmd loaded, $(if_changed ) will always arrange > for it to get re-built, because it then will consider the command > used to build the file to have changed (as the stored one appears to > be empty). Acked-by: Roger Pau Monné <roger.pau@citrix.com> Thanks, Roger.
On 15.09.2020 15:46, Roger Pau Monné wrote: > On Tue, Sep 15, 2020 at 02:26:34PM +0200, Jan Beulich wrote: >> On 15.09.2020 13:56, Roger Pau Monné wrote: >>> On Mon, Sep 14, 2020 at 12:15:39PM +0200, Jan Beulich wrote: >>>> Switch to $(call if_changed,ld) where possible; presumably not doing so >>>> in e321576f4047 ("xen/build: start using if_changed") right away was an >>>> oversight, as it did for Arm in (just) one case. It failed to add >>>> prelink.o to $(targets), though, causing - judging from the observed >>>> behavior on x86 - undue rebuilds of the final binary (because of >>>> prelink.o getting rebuild for $(cmd_prelink.o) being empty, in turn >>>> because of .prelink.o.cmd not getting read) during "make install-xen". >>> >>> I'm not sure I follow why prelink.o needs to be added to targets, does >>> this offer some kind of protection against rebuilds when doing make >>> install? >> >> In a way, but (as I view it) not really. It is the use of ... >> >>> The switch to if_changed LGTM. >> >> ... if_changed which requires this. .*.cmd files will only be loaded >> for anything explicitly or implicitly listed as a target. While .o >> coming from $(obj-y) get added there automatically, prelink.o is not >> something that could be recognized as needing adding, hence the >> "manual" insertion. > > This seems very prone to mistakes, as you have to remember that > whatever object that uses if_changed should also be part of the > targets set, or else it will get rebuild unconditionally. The use of $(if_changed ...) has further similar pitfalls: One may not forget to add FORCE to the dependencies, and there may not be blanks after the comma (where one would usually put one). But I think the benefits of this construct are so significant that this extra care that's needed is well justified. Plus this is actually mentioned (at least in passing) in docs/misc/xen-makefiles/makefile.rst. > I think adding some of the above reasoning to the commit message would > be helpful IMO. Well, I've merely re-explained what I think the commit message already says. >> Without .prelink.o.cmd loaded, $(if_changed ) will always arrange >> for it to get re-built, because it then will consider the command >> used to build the file to have changed (as the stored one appears to >> be empty). > > Acked-by: Roger Pau Monné <roger.pau@citrix.com> Thanks. Jan
On 14.09.2020 12:15, Jan Beulich wrote: > Switch to $(call if_changed,ld) where possible; presumably not doing so > in e321576f4047 ("xen/build: start using if_changed") right away was an > oversight, as it did for Arm in (just) one case. It failed to add > prelink.o to $(targets), though, causing - judging from the observed > behavior on x86 - undue rebuilds of the final binary (because of > prelink.o getting rebuild for $(cmd_prelink.o) being empty, in turn > because of .prelink.o.cmd not getting read) during "make install-xen". > > Signed-off-by: Jan Beulich <jbeulich@suse.com> > --- > xen/arch/arm/Makefile | 4 +++- > xen/arch/x86/Makefile | 18 ++++++++++-------- > 2 files changed, 13 insertions(+), 9 deletions(-) May I ask for an Arm-side ack (or otherwise) here, please? Jan > diff --git a/xen/arch/arm/Makefile b/xen/arch/arm/Makefile > index 51173d97127e..296c5e68bbc3 100644 > --- a/xen/arch/arm/Makefile > +++ b/xen/arch/arm/Makefile > @@ -95,12 +95,14 @@ prelink_lto.o: $(ALL_OBJS) > > # Link it with all the binary objects > prelink.o: $(patsubst %/built_in.o,%/built_in_bin.o,$(ALL_OBJS)) prelink_lto.o > - $(LD) $(XEN_LDFLAGS) -r -o $@ $^ > + $(call if_changed,ld) > else > prelink.o: $(ALL_OBJS) FORCE > $(call if_changed,ld) > endif > > +targets += prelink.o > + > $(TARGET)-syms: prelink.o xen.lds > $(LD) $(XEN_LDFLAGS) -T xen.lds -N prelink.o \ > $(BASEDIR)/common/symbols-dummy.o -o $(@D)/.$(@F).0 > diff --git a/xen/arch/x86/Makefile b/xen/arch/x86/Makefile > index 74152f2a0dad..9b368632fb43 100644 > --- a/xen/arch/x86/Makefile > +++ b/xen/arch/x86/Makefile > @@ -136,19 +136,21 @@ prelink_lto.o: $(ALL_OBJS) > $(LD_LTO) -r -o $@ $^ > > # Link it with all the binary objects > -prelink.o: $(patsubst %/built_in.o,%/built_in_bin.o,$(ALL_OBJS)) prelink_lto.o $(EFI_OBJS-y) > - $(LD) $(XEN_LDFLAGS) -r -o $@ $^ > +prelink.o: $(patsubst %/built_in.o,%/built_in_bin.o,$(ALL_OBJS)) prelink_lto.o $(EFI_OBJS-y) FORCE > + $(call if_changed,ld) > > -prelink-efi.o: $(patsubst %/built_in.o,%/built_in_bin.o,$(ALL_OBJS)) prelink_lto.o > - $(LD) $(XEN_LDFLAGS) -r -o $@ $^ > +prelink-efi.o: $(patsubst %/built_in.o,%/built_in_bin.o,$(ALL_OBJS)) prelink_lto.o FORCE > + $(call if_changed,ld) > else > -prelink.o: $(ALL_OBJS) $(EFI_OBJS-y) > - $(LD) $(XEN_LDFLAGS) -r -o $@ $^ > +prelink.o: $(ALL_OBJS) $(EFI_OBJS-y) FORCE > + $(call if_changed,ld) > > -prelink-efi.o: $(ALL_OBJS) > - $(LD) $(XEN_LDFLAGS) -r -o $@ $^ > +prelink-efi.o: $(ALL_OBJS) FORCE > + $(call if_changed,ld) > endif > > +targets += prelink.o prelink-efi.o > + > $(TARGET)-syms: prelink.o xen.lds > $(LD) $(XEN_LDFLAGS) -T xen.lds -N prelink.o $(build_id_linker) \ > $(BASEDIR)/common/symbols-dummy.o -o $(@D)/.$(@F).0 >
Hi Jan, On 21/09/2020 11:17, Jan Beulich wrote: > On 14.09.2020 12:15, Jan Beulich wrote: >> Switch to $(call if_changed,ld) where possible; presumably not doing so >> in e321576f4047 ("xen/build: start using if_changed") right away was an >> oversight, as it did for Arm in (just) one case. It failed to add >> prelink.o to $(targets), though, causing - judging from the observed >> behavior on x86 - undue rebuilds of the final binary (because of >> prelink.o getting rebuild for $(cmd_prelink.o) being empty, in turn >> because of .prelink.o.cmd not getting read) during "make install-xen". >> >> Signed-off-by: Jan Beulich <jbeulich@suse.com> >> --- >> xen/arch/arm/Makefile | 4 +++- >> xen/arch/x86/Makefile | 18 ++++++++++-------- >> 2 files changed, 13 insertions(+), 9 deletions(-) > > May I ask for an Arm-side ack (or otherwise) here, please? Acked-by: Julien Grall <jgrall@amazon.com> Cheers, > > Jan > >> diff --git a/xen/arch/arm/Makefile b/xen/arch/arm/Makefile >> index 51173d97127e..296c5e68bbc3 100644 >> --- a/xen/arch/arm/Makefile >> +++ b/xen/arch/arm/Makefile >> @@ -95,12 +95,14 @@ prelink_lto.o: $(ALL_OBJS) >> >> # Link it with all the binary objects >> prelink.o: $(patsubst %/built_in.o,%/built_in_bin.o,$(ALL_OBJS)) prelink_lto.o >> - $(LD) $(XEN_LDFLAGS) -r -o $@ $^ >> + $(call if_changed,ld) >> else >> prelink.o: $(ALL_OBJS) FORCE >> $(call if_changed,ld) >> endif >> >> +targets += prelink.o >> + >> $(TARGET)-syms: prelink.o xen.lds >> $(LD) $(XEN_LDFLAGS) -T xen.lds -N prelink.o \ >> $(BASEDIR)/common/symbols-dummy.o -o $(@D)/.$(@F).0 >> diff --git a/xen/arch/x86/Makefile b/xen/arch/x86/Makefile >> index 74152f2a0dad..9b368632fb43 100644 >> --- a/xen/arch/x86/Makefile >> +++ b/xen/arch/x86/Makefile >> @@ -136,19 +136,21 @@ prelink_lto.o: $(ALL_OBJS) >> $(LD_LTO) -r -o $@ $^ >> >> # Link it with all the binary objects >> -prelink.o: $(patsubst %/built_in.o,%/built_in_bin.o,$(ALL_OBJS)) prelink_lto.o $(EFI_OBJS-y) >> - $(LD) $(XEN_LDFLAGS) -r -o $@ $^ >> +prelink.o: $(patsubst %/built_in.o,%/built_in_bin.o,$(ALL_OBJS)) prelink_lto.o $(EFI_OBJS-y) FORCE >> + $(call if_changed,ld) >> >> -prelink-efi.o: $(patsubst %/built_in.o,%/built_in_bin.o,$(ALL_OBJS)) prelink_lto.o >> - $(LD) $(XEN_LDFLAGS) -r -o $@ $^ >> +prelink-efi.o: $(patsubst %/built_in.o,%/built_in_bin.o,$(ALL_OBJS)) prelink_lto.o FORCE >> + $(call if_changed,ld) >> else >> -prelink.o: $(ALL_OBJS) $(EFI_OBJS-y) >> - $(LD) $(XEN_LDFLAGS) -r -o $@ $^ >> +prelink.o: $(ALL_OBJS) $(EFI_OBJS-y) FORCE >> + $(call if_changed,ld) >> >> -prelink-efi.o: $(ALL_OBJS) >> - $(LD) $(XEN_LDFLAGS) -r -o $@ $^ >> +prelink-efi.o: $(ALL_OBJS) FORCE >> + $(call if_changed,ld) >> endif >> >> +targets += prelink.o prelink-efi.o >> + >> $(TARGET)-syms: prelink.o xen.lds >> $(LD) $(XEN_LDFLAGS) -T xen.lds -N prelink.o $(build_id_linker) \ >> $(BASEDIR)/common/symbols-dummy.o -o $(@D)/.$(@F).0 >> >
On 21.09.2020 13:39, Julien Grall wrote: > On 21/09/2020 11:17, Jan Beulich wrote: >> On 14.09.2020 12:15, Jan Beulich wrote: >>> Switch to $(call if_changed,ld) where possible; presumably not doing so >>> in e321576f4047 ("xen/build: start using if_changed") right away was an >>> oversight, as it did for Arm in (just) one case. It failed to add >>> prelink.o to $(targets), though, causing - judging from the observed >>> behavior on x86 - undue rebuilds of the final binary (because of >>> prelink.o getting rebuild for $(cmd_prelink.o) being empty, in turn >>> because of .prelink.o.cmd not getting read) during "make install-xen". >>> >>> Signed-off-by: Jan Beulich <jbeulich@suse.com> >>> --- >>> xen/arch/arm/Makefile | 4 +++- >>> xen/arch/x86/Makefile | 18 ++++++++++-------- >>> 2 files changed, 13 insertions(+), 9 deletions(-) >> >> May I ask for an Arm-side ack (or otherwise) here, please? > > Acked-by: Julien Grall <jgrall@amazon.com> Thanks. On the Arm side this is actually addressing a (minor) bug, so I wonder whether I should queue this up for backporting. Do you have an opinion either way? Jan
Hi Jan, On 22/09/2020 09:28, Jan Beulich wrote: > On 21.09.2020 13:39, Julien Grall wrote: >> On 21/09/2020 11:17, Jan Beulich wrote: >>> On 14.09.2020 12:15, Jan Beulich wrote: >>>> Switch to $(call if_changed,ld) where possible; presumably not doing so >>>> in e321576f4047 ("xen/build: start using if_changed") right away was an >>>> oversight, as it did for Arm in (just) one case. It failed to add >>>> prelink.o to $(targets), though, causing - judging from the observed >>>> behavior on x86 - undue rebuilds of the final binary (because of >>>> prelink.o getting rebuild for $(cmd_prelink.o) being empty, in turn >>>> because of .prelink.o.cmd not getting read) during "make install-xen". >>>> >>>> Signed-off-by: Jan Beulich <jbeulich@suse.com> >>>> --- >>>> xen/arch/arm/Makefile | 4 +++- >>>> xen/arch/x86/Makefile | 18 ++++++++++-------- >>>> 2 files changed, 13 insertions(+), 9 deletions(-) >>> >>> May I ask for an Arm-side ack (or otherwise) here, please? >> >> Acked-by: Julien Grall <jgrall@amazon.com> > > Thanks. On the Arm side this is actually addressing a (minor) bug, Just to confirm, the bug is: Xen will be rebuilt when it is not necessary, right? Cheers,
On 22.09.2020 11:24, Julien Grall wrote: > On 22/09/2020 09:28, Jan Beulich wrote: >> On 21.09.2020 13:39, Julien Grall wrote: >>> On 21/09/2020 11:17, Jan Beulich wrote: >>>> On 14.09.2020 12:15, Jan Beulich wrote: >>>>> Switch to $(call if_changed,ld) where possible; presumably not doing so >>>>> in e321576f4047 ("xen/build: start using if_changed") right away was an >>>>> oversight, as it did for Arm in (just) one case. It failed to add >>>>> prelink.o to $(targets), though, causing - judging from the observed >>>>> behavior on x86 - undue rebuilds of the final binary (because of >>>>> prelink.o getting rebuild for $(cmd_prelink.o) being empty, in turn >>>>> because of .prelink.o.cmd not getting read) during "make install-xen". >>>>> >>>>> Signed-off-by: Jan Beulich <jbeulich@suse.com> >>>>> --- >>>>> xen/arch/arm/Makefile | 4 +++- >>>>> xen/arch/x86/Makefile | 18 ++++++++++-------- >>>>> 2 files changed, 13 insertions(+), 9 deletions(-) >>>> >>>> May I ask for an Arm-side ack (or otherwise) here, please? >>> >>> Acked-by: Julien Grall <jgrall@amazon.com> >> >> Thanks. On the Arm side this is actually addressing a (minor) bug, > > Just to confirm, the bug is: Xen will be rebuilt when it is not > necessary, right? Yes. When building as non-root but installing as root, this would typically involve an owner change of some of the involved files. That's how I did notice the issue on x86 (after switching to if_changed) in the first place. Jan
Hi Jan, On 22/09/2020 11:55, Jan Beulich wrote: > On 22.09.2020 11:24, Julien Grall wrote: >> On 22/09/2020 09:28, Jan Beulich wrote: >>> On 21.09.2020 13:39, Julien Grall wrote: >>>> On 21/09/2020 11:17, Jan Beulich wrote: >>>>> On 14.09.2020 12:15, Jan Beulich wrote: >>>>>> Switch to $(call if_changed,ld) where possible; presumably not doing so >>>>>> in e321576f4047 ("xen/build: start using if_changed") right away was an >>>>>> oversight, as it did for Arm in (just) one case. It failed to add >>>>>> prelink.o to $(targets), though, causing - judging from the observed >>>>>> behavior on x86 - undue rebuilds of the final binary (because of >>>>>> prelink.o getting rebuild for $(cmd_prelink.o) being empty, in turn >>>>>> because of .prelink.o.cmd not getting read) during "make install-xen". >>>>>> >>>>>> Signed-off-by: Jan Beulich <jbeulich@suse.com> >>>>>> --- >>>>>> xen/arch/arm/Makefile | 4 +++- >>>>>> xen/arch/x86/Makefile | 18 ++++++++++-------- >>>>>> 2 files changed, 13 insertions(+), 9 deletions(-) >>>>> >>>>> May I ask for an Arm-side ack (or otherwise) here, please? >>>> >>>> Acked-by: Julien Grall <jgrall@amazon.com> >>> >>> Thanks. On the Arm side this is actually addressing a (minor) bug, >> >> Just to confirm, the bug is: Xen will be rebuilt when it is not >> necessary, right? > > Yes. When building as non-root but installing as root, this would > typically involve an owner change of some of the involved files. > That's how I did notice the issue on x86 (after switching to > if_changed) in the first place. Thanks for the explanation. I think it would be fine to backport. @Stefano, what do you think? Cheers,
On Tue, 22 Sep 2020, Julien Grall wrote: > Hi Jan, > > On 22/09/2020 11:55, Jan Beulich wrote: > > On 22.09.2020 11:24, Julien Grall wrote: > > > On 22/09/2020 09:28, Jan Beulich wrote: > > > > On 21.09.2020 13:39, Julien Grall wrote: > > > > > On 21/09/2020 11:17, Jan Beulich wrote: > > > > > > On 14.09.2020 12:15, Jan Beulich wrote: > > > > > > > Switch to $(call if_changed,ld) where possible; presumably not > > > > > > > doing so > > > > > > > in e321576f4047 ("xen/build: start using if_changed") right away > > > > > > > was an > > > > > > > oversight, as it did for Arm in (just) one case. It failed to add > > > > > > > prelink.o to $(targets), though, causing - judging from the > > > > > > > observed > > > > > > > behavior on x86 - undue rebuilds of the final binary (because of > > > > > > > prelink.o getting rebuild for $(cmd_prelink.o) being empty, in > > > > > > > turn > > > > > > > because of .prelink.o.cmd not getting read) during "make > > > > > > > install-xen". > > > > > > > > > > > > > > Signed-off-by: Jan Beulich <jbeulich@suse.com> > > > > > > > --- > > > > > > > xen/arch/arm/Makefile | 4 +++- > > > > > > > xen/arch/x86/Makefile | 18 ++++++++++-------- > > > > > > > 2 files changed, 13 insertions(+), 9 deletions(-) > > > > > > > > > > > > May I ask for an Arm-side ack (or otherwise) here, please? > > > > > > > > > > Acked-by: Julien Grall <jgrall@amazon.com> > > > > > > > > Thanks. On the Arm side this is actually addressing a (minor) bug, > > > > > > Just to confirm, the bug is: Xen will be rebuilt when it is not > > > necessary, right? > > > > Yes. When building as non-root but installing as root, this would > > typically involve an owner change of some of the involved files. > > That's how I did notice the issue on x86 (after switching to > > if_changed) in the first place. > > Thanks for the explanation. I think it would be fine to backport. > @Stefano, what do you think? I am OK with that
diff --git a/xen/arch/arm/Makefile b/xen/arch/arm/Makefile index 51173d97127e..296c5e68bbc3 100644 --- a/xen/arch/arm/Makefile +++ b/xen/arch/arm/Makefile @@ -95,12 +95,14 @@ prelink_lto.o: $(ALL_OBJS) # Link it with all the binary objects prelink.o: $(patsubst %/built_in.o,%/built_in_bin.o,$(ALL_OBJS)) prelink_lto.o - $(LD) $(XEN_LDFLAGS) -r -o $@ $^ + $(call if_changed,ld) else prelink.o: $(ALL_OBJS) FORCE $(call if_changed,ld) endif +targets += prelink.o + $(TARGET)-syms: prelink.o xen.lds $(LD) $(XEN_LDFLAGS) -T xen.lds -N prelink.o \ $(BASEDIR)/common/symbols-dummy.o -o $(@D)/.$(@F).0 diff --git a/xen/arch/x86/Makefile b/xen/arch/x86/Makefile index 74152f2a0dad..9b368632fb43 100644 --- a/xen/arch/x86/Makefile +++ b/xen/arch/x86/Makefile @@ -136,19 +136,21 @@ prelink_lto.o: $(ALL_OBJS) $(LD_LTO) -r -o $@ $^ # Link it with all the binary objects -prelink.o: $(patsubst %/built_in.o,%/built_in_bin.o,$(ALL_OBJS)) prelink_lto.o $(EFI_OBJS-y) - $(LD) $(XEN_LDFLAGS) -r -o $@ $^ +prelink.o: $(patsubst %/built_in.o,%/built_in_bin.o,$(ALL_OBJS)) prelink_lto.o $(EFI_OBJS-y) FORCE + $(call if_changed,ld) -prelink-efi.o: $(patsubst %/built_in.o,%/built_in_bin.o,$(ALL_OBJS)) prelink_lto.o - $(LD) $(XEN_LDFLAGS) -r -o $@ $^ +prelink-efi.o: $(patsubst %/built_in.o,%/built_in_bin.o,$(ALL_OBJS)) prelink_lto.o FORCE + $(call if_changed,ld) else -prelink.o: $(ALL_OBJS) $(EFI_OBJS-y) - $(LD) $(XEN_LDFLAGS) -r -o $@ $^ +prelink.o: $(ALL_OBJS) $(EFI_OBJS-y) FORCE + $(call if_changed,ld) -prelink-efi.o: $(ALL_OBJS) - $(LD) $(XEN_LDFLAGS) -r -o $@ $^ +prelink-efi.o: $(ALL_OBJS) FORCE + $(call if_changed,ld) endif +targets += prelink.o prelink-efi.o + $(TARGET)-syms: prelink.o xen.lds $(LD) $(XEN_LDFLAGS) -T xen.lds -N prelink.o $(build_id_linker) \ $(BASEDIR)/common/symbols-dummy.o -o $(@D)/.$(@F).0
Switch to $(call if_changed,ld) where possible; presumably not doing so in e321576f4047 ("xen/build: start using if_changed") right away was an oversight, as it did for Arm in (just) one case. It failed to add prelink.o to $(targets), though, causing - judging from the observed behavior on x86 - undue rebuilds of the final binary (because of prelink.o getting rebuild for $(cmd_prelink.o) being empty, in turn because of .prelink.o.cmd not getting read) during "make install-xen". Signed-off-by: Jan Beulich <jbeulich@suse.com> --- xen/arch/arm/Makefile | 4 +++- xen/arch/x86/Makefile | 18 ++++++++++-------- 2 files changed, 13 insertions(+), 9 deletions(-)