tools build: Unused function, incomplete rename
Subsystems (3)
Arnaldo Carvalho de Melo
maintainerHi Jiri, Ingo,
While trying to figure out why the bpf feature test is always
triggering the display of the "Auto-detecting system features" I noticed this pattern:
[acme@felicio linux]$ egrep '^define|eval' tools/build/Makefile.feature feature_check = $(eval $(feature_check_code))
define feature_check_codefeature_set = $(eval $(feature_set_code))
define feature_set_codeset_test_all_flags = $(eval $(set_test_all_flags_code))
define set_test_all_flags_codefeature_print_status = $(eval $(feature_print_status_code)) $(info $(MSG)) define feature_print_status_codefeature_print_text = $(eval $(feature_print_text_code)) $(info $(MSG)) define feature_print_text_codefeature_display_check = $(eval $(feature_check_code))
define feature_display_check_code
[acme@felicio linux]$
In all but one case the eval matches the following define, except for the last one, don't we need the following patch?
$ git diff tools/build/Makefile.feature
tools/build/Makefile.feature +1 -1
@@ -142,7 +142,7 @@ ifneq ("$(FEATURE_DUMP)","$(FEATURE_DUMP_FILE)") feature_display := 1 endif -feature_display_check = $(eval $(feature_check_code)) +feature_display_check = $(eval $(feature_check_display_code)) define feature_display_check_code ifneq ($(feature-$(1)), 1) feature_display := 1
---------------------------------------------------------------------
I guess the bug was introduced here:
commit 58d4f00ff13f20468f8fa8edcb57a195c31af46d
Author: Jiri Olsa <[email protected]> Date: Thu Mar 19 20:48:49 2015 +0100
perf build: Fix feature_check name clash
We have 2 feature_check functions, which conflict with each other. Fixing it by renaming the latter to feature_display_check.
tools/perf/config/Makefile +3 -3
@@ -805,14 +805,14 @@ ifneq ("$(FEATURE_DUMP)","$(FEATURE_DUMP_FILE)") feature_display := 1 endif -feature_check = $(eval $(feature_check_code)) -define feature_check_code +feature_display_check = $(eval $(feature_check_code)) +define feature_display_check_code ifneq ($(feature-$(1)), 1) feature_display := 1 endif endef -$(foreach feat,$(FEATURE_DISPLAY),$(call feature_check,$(feat))) +$(foreach feat,$(FEATURE_DISPLAY),$(call feature_display_check,$(feat)))
Arnaldo Carvalho de Melo
maintainerEm Fri, Sep 18, 2015 at 04:23:34PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 5 lines
Hi Jiri, Ingo,
While trying to figure out why the bpf feature test is always
triggering the display of the "Auto-detecting system features" I noticed this pattern:
Another problem, this time in how tools/lib/bpf/ specifies which features to test for and which ones should have the feature detection shown, does the following patch makes sense? I think it does because FEATURE_TESTS looks like the ones that will be tested, and FEATURE_DISPLAY the ones that will appear...:
tools/lib/bpf/Makefile +2 -2
@@ -64,8 +64,8 @@ srctree := $(patsubst %/,%,$(dir $(srctree))) #$(info Determined 'srctree' to be $(srctree)) endif -FEATURE_DISPLAY = libelf libelf-getphdrnum libelf-mmap bpf -FEATURE_TESTS = libelf bpf +FEATURE_TESTS = libelf libelf-getphdrnum libelf-mmap bpf +FEATURE_DISPLAY = libelf bpf INCLUDES = -I. -I$(srctree)/tools/include -I$(srctree)/arch/$(ARCH)/include/uapi -I$(srctree)/include/uapi FEATURE_CHECK_CFLAGS-bpf = $(INCLUDES)
Arnaldo Carvalho de Melo
maintainerEm Fri, Sep 18, 2015 at 04:38:42PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 10 lines
Em Fri, Sep 18, 2015 at 04:23:34PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 3 lines
While trying to figure out why the bpf feature test is always
triggering the display of the "Auto-detecting system features" I noticed this pattern:Another problem, this time in how
tools/lib/bpf/specifies which features to test for and which ones should have the feature detection shown, does the following patch makes sense? I think it does becauseFEATURE_TESTSlooks like the ones that will be tested, andFEATURE_DISPLAYthe ones that will appear...:
So the original problem seems to be this:
[acme@felicio linux]$ cat /tmp/build/perf/FEATURE-DUMP feature-libelf(1) feature-libelf-getphdrnum(1) feature-libelf-mmap(1) feature-bpf(1)
This is the content at the end of a build, i.e. the FEATURE-DUMP for tools/lib/ebpf/ usage of the feature detection system, since tools/perf/ uses the same file and selects a different set of features.
I think that ebpf should use a separate directory, inside $(OUTPUT), this way we would have $(OUTPUT)/FEATURE-DUMP for perf and $(OUTPUT)/bpf/FEATURE-DUMP for ebpf.
- Arnaldo
Arnaldo Carvalho de Melo
maintainerEm Fri, Sep 18, 2015 at 04:48:52PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 17 lines
Em Fri, Sep 18, 2015 at 04:38:42PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 10 lines
Em Fri, Sep 18, 2015 at 04:23:34PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 3 lines
While trying to figure out why the bpf feature test is always
triggering the display of the "Auto-detecting system features" I noticed this pattern:Another problem, this time in how
tools/lib/bpf/specifies which features to test for and which ones should have the feature detection shown, does the following patch makes sense? I think it does becauseFEATURE_TESTSlooks like the ones that will be tested, andFEATURE_DISPLAYthe ones that will appear...:So the original problem seems to be this:
[acme@felicio linux]$ cat
/tmp/build/perf/FEATURE-DUMPfeature-libelf(1) feature-libelf-getphdrnum(1) feature-libelf-mmap(1) feature-bpf(1)
Quoted 8 lines
This is the content at the end of a build, i.e. the FEATURE-DUMP for
tools/lib/ebpf/usage of the feature detection system, sincetools/perf/uses the same file and selects a different set of features.I think that ebpf should use a separate directory, inside $(OUTPUT), this way we would have $(OUTPUT)/FEATURE-DUMP for perf and $(OUTPUT)/bpf/FEATURE-DUMP for ebpf.
[acme@felicio linux]$ ls -la /tmp/build/perf/FEATURE-DUMP -rw-rw-r--. 1 acme acme 338 Sep 18 17:38 /tmp/build/perf/FEATURE-DUMP [acme@felicio linux]$ ls -la /tmp/build/perf/FEATURE-DUMP.libbpf -rw-rw-r--. 1 acme acme 85 Sep 18 17:38 /tmp/build/perf/FEATURE-DUMP.libbpf [acme@felicio linux]$
Ok, patch below fixes this one, now the second run doesn't auto detects things again, i.e. libbpf feature detection doesn't stomps on perf's, I'll get those patches in a patchkit and send over the weekend. If you find anything fishy with it, holler.
- Arnaldo
tools/build/Makefile.feature +3 -2
@@ -123,8 +123,9 @@ define feature_print_text_code MSG = $(shell printf '...%30s: %s' $(1) $(2)) endef +FEATURE_DUMP_FILENAME = $(OUTPUT)FEATURE-DUMP$(FEATURE_USER) FEATURE_DUMP := $(foreach feat,$(FEATURE_DISPLAY),feature-$(feat)($(feature-$(feat)))) -FEATURE_DUMP_FILE := $(shell touch $(OUTPUT)FEATURE-DUMP; cat $(OUTPUT)FEATURE-DUMP) +FEATURE_DUMP_FILE := $(shell touch $(FEATURE_DUMP_FILENAME); cat $(FEATURE_DUMP_FILENAME)) ifeq ($(dwarf-post-unwind),1) FEATURE_DUMP += dwarf-post-unwind($(dwarf-post-unwind-text))
@@ -138,6 +139,6 @@ endif # - VF is enabled ifneq ("$(FEATURE_DUMP)","$(FEATURE_DUMP_FILE)") - $(shell echo "$(FEATURE_DUMP)" > $(OUTPUT)FEATURE-DUMP) + $(shell echo "$(FEATURE_DUMP)" > $(FEATURE_DUMP_FILENAME)) feature_display := 1 endif
tools/lib/bpf/Makefile +1 -0
@@ -64,8 +64,8 @@ srctree := $(patsubst %/,%,$(dir $(srctree))) #$(info Determined 'srctree' to be $(srctree)) endif +FEATURE_USER = .libbpf FEATURE_DISPLAY = libelf libelf-getphdrnum libelf-mmap bpf FEATURE_TESTS = libelf bpf INCLUDES = -I. -I$(srctree)/tools/include -I$(srctree)/arch/$(ARCH)/include/uapi -I$(srctree)/include/uapi
FEATURE_CHECK_CFLAGS-bpf = $(INCLUDES)
Quoted 33 lines · Arnaldo Carvalho de Melo
Em Fri, Sep 18, 2015 at 04:48:52PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 17 lines
Em Fri, Sep 18, 2015 at 04:38:42PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 10 lines
Em Fri, Sep 18, 2015 at 04:23:34PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 3 lines
While trying to figure out why the bpf feature test is always
triggering the display of the "Auto-detecting system features" I noticed this pattern:Another problem, this time in how
tools/lib/bpf/specifies which features to test for and which ones should have the feature detection shown, does the following patch makes sense? I think it does becauseFEATURE_TESTSlooks like the ones that will be tested, andFEATURE_DISPLAYthe ones that will appear...:So the original problem seems to be this:
[acme@felicio linux]$ cat
/tmp/build/perf/FEATURE-DUMPfeature-libelf(1) feature-libelf-getphdrnum(1) feature-libelf-mmap(1) feature-bpf(1)Quoted 8 lines
This is the content at the end of a build, i.e. the FEATURE-DUMP for
tools/lib/ebpf/usage of the feature detection system, sincetools/perf/uses the same file and selects a different set of features.I think that ebpf should use a separate directory, inside $(OUTPUT), this way we would have $(OUTPUT)/FEATURE-DUMP for perf and $(OUTPUT)/bpf/FEATURE-DUMP for ebpf.
[acme@felicio linux]$ ls -la
/tmp/build/perf/FEATURE-DUMP-rw-rw-r--. 1 acme acme 338 Sep 18 17:38/tmp/build/perf/FEATURE-DUMP[acme@felicio linux]$ ls -la/tmp/build/perf/FEATURE-DUMP.libbpf-rw-rw-r--. 1 acme acme 85 Sep 18 17:38/tmp/build/perf/FEATURE-DUMP.libbpf[acme@felicio linux]$
Quoted 4 lines
Ok, patch below fixes this one, now the second run doesn't auto detects things again, i.e. libbpf feature detection doesn't stomps on perf's, I'll get those patches in a patchkit and send over the weekend. If you find anything fishy with it, holler.
it looks ok.. btw IMO sharing output directory for perf and related libs could bite us in the future.. I think we should base it to tools directory.. I'll try to come up with something
thanks,
jirka
Quoted 31 lines · Arnaldo Carvalho de Melo
Hi Jiri, Ingo,
While trying to figure out why the bpf feature test is always
triggering the display of the "Auto-detecting system features" I noticed this pattern:[acme@felicio linux]$ egrep '^define|eval'
tools/build/Makefile.featurefeature_check= $(eval $(feature_check_code))
definefeature_check_codefeature_set= $(eval $(feature_set_code))
definefeature_set_codeset_test_all_flags= $(eval $(set_test_all_flags_code))
defineset_test_all_flags_codefeature_print_status= $(eval $(feature_print_status_code)) $(info $(MSG)) definefeature_print_status_codefeature_print_text= $(eval $(feature_print_text_code)) $(info $(MSG)) definefeature_print_text_codefeature_display_check= $(eval $(feature_check_code))
definefeature_display_check_code
[acme@felicio linux]$In all but one case the eval matches the following define, except for the last one, don't we need the following patch?
$ git diff tools/build/Makefile.featuretools/build/Makefile.feature +0
-0
tools/build/Makefile.feature ifneq ("$(FEATURE_DUMP)","$(FEATURE_DUMP_FILE)") · line 148
Quoted 8 lines
tools/build/Makefile.feature +1
-1@@ -142,7 +142,7 @@ ifneq ("$(FEATURE_DUMP)","$(FEATURE_DUMP_FILE)") feature_display := 1 endif -feature_display_check = $(eval $(feature_check_code)) +feature_display_check = $(eval $(feature_check_display_code)) define feature_display_check_code ifneq ($(feature-$(1)), 1) feature_display := 1
ouch, that is a bug.. nice catch!
thanks,
jirka