tools/libs/light: fix race in Makefile
PATCH 1/1 tools/libs/light: fix race in Makefile
The header $(INCLUDE)/_lixl_list.h matches two different rules, which can result in build breakage. Fix that.
--- tools/libs/light/Makefile | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
tools/libs/light/Makefile +3 -3
@@ -203,9 +203,9 @@ _libxl.api-for-check: $(XEN_INCLUDE)/libxl.h $(AUTOINCS) >[email protected] mv -f [email protected] $@ -$(XEN_INCLUDE)/_libxl_list.h: $(XEN_INCLUDE)/xen-external/bsd-sys-queue-h-seddery $(XEN_INCLUDE)/xen-external/bsd-sys-queue.h - $(PERL) $^ --prefix=libxl >$(notdir $@).new - $(call move-if-changed,$(notdir $@).new,$@) +_libxl_list.h: $(XEN_INCLUDE)/xen-external/bsd-sys-queue-h-seddery $(XEN_INCLUDE)/xen-external/bsd-sys-queue.h + $(PERL) $^ --prefix=libxl >[email protected] + $(call move-if-changed,[email protected],$@) _libxl_save_msgs_helper.c _libxl_save_msgs_callout.c \ _libxl_save_msgs_helper.h _libxl_save_msgs_callout.h: \
Signature · 1 line
2.26.2
Quoted 2 lines · Juergen Gross
The header $(INCLUDE)/_lixl_list.h matches two different rules, which can result in build breakage. Fix that.
While I don't doubt you having observed a race, I'm not sure this is true, and hence I'm also not sure the change is going to address it:
Aiui the two rules you talk about are the one you change and
$(XEN_INCLUDE)/_%.h: _%.h
$(call move-if-changed,_$*.h,$(XEN_INCLUDE)/_$*.h)
But a pattern rule doesn't come into play when a specific rule for a file exists.
What I don't understand here is why this two step moving around of headers is used: Instead of the above pattern rule, can't the rule to generate _libxl_type%.h, _libxl_type%_json.h, _libxl_type%_private.h, and _libxl_type%.c put the relevant header files right into their designated place? This would allow the pattern rule to go away, albeit I'd then still be unclear about the specific race you did observe.
Jan
Quoted 23 lines
Signed-off-by Juergen Gross <[email protected]>--- tools/libs/light/Makefile | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)tools/libs/light/Makefile +3
-3@@ -203,9 +203,9 @@ _libxl.api-for-check: $(XEN_INCLUDE)/libxl.h $(AUTOINCS) >[email protected] mv -f [email protected] $@ -$(XEN_INCLUDE)/_libxl_list.h: $(XEN_INCLUDE)/xen-external/bsd-sys-queue-h-seddery $(XEN_INCLUDE)/xen-external/bsd-sys-queue.h - $(PERL) $^ --prefix=libxl >$(notdir $@).new - $(call move-if-changed,$(notdir $@).new,$@) +_libxl_list.h: $(XEN_INCLUDE)/xen-external/bsd-sys-queue-h-seddery $(XEN_INCLUDE)/xen-external/bsd-sys-queue.h + $(PERL) $^ --prefix=libxl >[email protected] + $(call move-if-changed,[email protected],$@) _libxl_save_msgs_helper.c _libxl_save_msgs_callout.c \ _libxl_save_msgs_helper.h _libxl_save_msgs_callout.h: \
Quoted 8 lines · Jan Beulich
Quoted 2 lines · Juergen Gross
The header $(INCLUDE)/_lixl_list.h matches two different rules, which can result in build breakage. Fix that.
While I don't doubt you having observed a race, I'm not sure this is true, and hence I'm also not sure the change is going to address it:
Aiui the two rules you talk about are the one you change and
Quoted 5 lines
$(XEN_INCLUDE)/_%.h: _%.h
$(call move-if-changed,_$*.h,$(XEN_INCLUDE)/_$*.h)But a pattern rule doesn't come into play when a specific rule for a file exists.
Hmm, true. I didn't see the race, but spotted the suspected ambiguity just by chance.
Quoted 8 lines
What I don't understand here is why this two step moving around of headers is used: Instead of the above pattern rule, can't the rule to generate _libxl_type%.h, _libxl_type%_json.h, _libxl_type%_private.h, and _libxl_type%.c put the relevant header files right into their designated place? This would allow the pattern rule to go away, albeit I'd then still be unclear about the specific race you did observe.
This would require to replace the pattern rules used to generate the files by per-file rules instead, as e.g. _libxl_types_json.h and _libxl_types_internal_json.h are matching the same pattern, but they need to end up in different directories.
In the end I think this patch can just be dropped.
Sorry for the noise,
Juergen
Quoted 9 lines · Jürgen Groß
Quoted 7 lines · Jan Beulich
What I don't understand here is why this two step moving around of headers is used: Instead of the above pattern rule, can't the rule to generate _libxl_type%.h, _libxl_type%_json.h, _libxl_type%_private.h, and _libxl_type%.c put the relevant header files right into their designated place? This would allow the pattern rule to go away, albeit I'd then still be unclear about the specific race you did observe.
Quoted 4 lines
This would require to replace the pattern rules used to generate the files by per-file rules instead, as e.g.
_libxl_types_json.hand_libxl_types_internal_json.hare matching the same pattern, but they need to end up in different directories.
Ah, right - I didn't pay attention to the *_internal*.h needs.
Jan