Specifying CFLAGS for a directory on the command line
Hello kbuild maintainers & list,
Years ago I used to be able to specify additional CFLAGS for a specific subdirectory on the command line, which I used for enabling gcov profiling without editing the subdirectory makefile, like so:
make GCOV_PROFILE_fs_bcachefs=y
But this stopped working ages ago, I believe in the 3.x era. Does anyone know if there's a more modern way to do this?
Cheers,
Kent
Quoted 4 lines · Kent Overstreet
Hello kbuild maintainers & list,
Years ago I used to be able to specify additional CFLAGS for a specific
Probably cause it's KCFLAGS ;)
I used this yesterday, it works.
Quoted 10 lines
subdirectory on the command line, which I used for enabling gcov profiling without editing the subdirectory makefile, like so:
make GCOV_PROFILE_fs_bcachefs=y
But this stopped working ages ago, I believe in the 3.x era. Does anyone know if there's a more modern way to do this?
Cheers,
Kent
Signature · 2 lines
Thanks, ~Nick Desaulniers
Quoted 4 lines · Nick Desaulniers
Quoted 1 line · Kent Overstreet
Quoted 9 lines
Quoted 3 lines
Hello kbuild maintainers & list,
Years ago I used to be able to specify additional CFLAGS for a specific
Probably cause it's KCFLAGS ;)
I used this yesterday, it works.
Quoted 1 line
subdirectory on the command line, which I used for enabling gcov
Ah, for a specific subdir? No I've not seen that, at least from the command line. Such flags can be specified via Makefile subdir-ccflags-y.
Quoted 15 lines
Quoted 9 lines
profiling without editing the subdirectory makefile, like so:
make GCOV_PROFILE_fs_bcachefs=y
But this stopped working ages ago, I believe in the 3.x era. Does anyone know if there's a more modern way to do this?
Cheers,
Kent--
Thanks,
~Nick Desaulniers
Signature · 2 lines
Thanks, ~Nick Desaulniers
Adding Peter to the cc, because I just realized gcov has a maintainer :)
Quoted 11 lines · Nick Desaulniers
Quoted 10 lines · Nick Desaulniers
Quoted 4 lines · Kent Overstreet
Hello kbuild maintainers & list,
Years ago I used to be able to specify additional CFLAGS for a specific
Probably cause it's KCFLAGS ;)
Quoted 7 lines
Quoted 3 lines
I used this yesterday, it works.
Quoted 1 line
subdirectory on the command line, which I used for enabling gcov
Ah, for a specific subdir? No I've not seen that, at least from the command line. Such flags can be specified via
Makefilesubdir-ccflags-y.
Sorry, not CFLAGS, I misread my old code - it's just a make variable.
From Documentation/dev-tools/gcov.rst, you enable gcov on a specific subdirectory by editing that directory's makefile, and adding
GCOV_PROFILE := y
or, for a specific file within that directory,
GCOV_PROFILE_main.o := y
So, if appending a file to GCOV_PROFILE works, why not a path?
This used to work - my old code would pass GCOV_PROFILE_fs_bcachefs=y on the make command line, but doesn't anymore.
Alas I have nowhere near the make-fu to debug this, and I believe I tried to bisect this back in the day but got nowhere... :)
Quoted 29 lines · Kent Overstreet
Adding Peter to the cc, because I just realized gcov has a maintainer :)
Quoted 11 lines · Nick Desaulniers
Quoted 10 lines · Nick Desaulniers
Quoted 4 lines · Kent Overstreet
Hello kbuild maintainers & list,
Years ago I used to be able to specify additional CFLAGS for a specific
Probably cause it's KCFLAGS ;)
Quoted 7 lines
Quoted 3 lines
I used this yesterday, it works.
Quoted 1 line
subdirectory on the command line, which I used for enabling gcov
Ah, for a specific subdir? No I've not seen that, at least from the command line. Such flags can be specified via
Makefilesubdir-ccflags-y.Sorry, not CFLAGS, I misread my old code - it's just a make variable.
From
Documentation/dev-tools/gcov.rst, you enable gcov on a specific subdirectory by editing that directory's makefile, and adding
GCOV_PROFILE:= y
Quoted 8 lines
or, for a specific file within that directory,
GCOV_PROFILE_main.o := y
So, if appending a file to
GCOV_PROFILEworks, why not a path?This used to work - my old code would pass GCOV_PROFILE_fs_bcachefs=y on the make command line, but doesn't anymore.
I'm unaware of any kbuild support for setting GCOV_PROFILE for a specific sub-directory from the command line, only from within the associated Makefile. I'm not sure how this could have worked in the past with the provide sample command line.
Here's how GCOV_PROFILE evaluation works (from scripts/Makefile.lib):
ifeq ($(CONFIG_GCOV_KERNEL),y)_c_flags += $(if $(patsubst n%,, \
$(GCOV_PROFILE_$(basetarget).o)$(GCOV_PROFILE)$(CONFIG_GCOV_PROFILE_ALL)),\
$(CFLAGS_GCOV))
endif
This bit of Makefile code determines whether to add the flags needed to enabled gcov profiling (CFLAGS_GCOV) to the compiler flags for the current compilation unit (_c_flags) by looking at the concatenation of the following variables:
- GCOV_PROFILE_<target base name>.o
- GCOV_PROFILE
- CONFIG_GCOV_PROFILE_ALL
gcov flags are only added if this concatenation does not start with an "n", and at least one of these variables is set to a non-empty value other than "n" ("y" typically). The "starts with" part is required to enable precedence for the more specific variable, e.g. an "n" in GCOV_PROFILE_filename.o overwrites a "y" in GCOV_PROFILE.
As you can see, there is no reference to a GCOV_PROFILE variable that is named after the sub-directory for which profiling should be enabled.
Signature · 2 lines
Peter Oberparleiter Linux on IBM Z Development - IBM Germany R&D
Quoted 22 lines · Peter Oberparleiter
I'm unaware of any kbuild support for setting
GCOV_PROFILEfor a specific sub-directory from the command line, only from within the associatedMakefile. I'm not sure how this could have worked in the past with the provide sample command line.Here's how
GCOV_PROFILEevaluation works (fromscripts/Makefile.lib):ifeq ($(CONFIG_GCOV_KERNEL),y)
_c_flags+= $(if $(patsubst n%,, \
$(GCOV_PROFILE_$(basetarget).o)$(GCOV_PROFILE)$(CONFIG_GCOV_PROFILE_ALL)),\
$(CFLAGS_GCOV))
endifThis bit of
Makefilecode determines whether to add the flags needed to enabled gcov profiling (CFLAGS_GCOV) to the compiler flags for the current compilation unit (_c_flags) by looking at the concatenation of the following variables:- GCOV_PROFILE_<target base name>.o
-GCOV_PROFILE
-CONFIG_GCOV_PROFILE_ALL
Quoted 8 lines
gcov flags are only added if this concatenation does not start with an "n", and at least one of these variables is set to a non-empty value other than "n" ("y" typically). The "starts with" part is required to enable precedence for the more specific variable, e.g. an "n" in GCOV_PROFILE_filename.o overwrites a "y" in
GCOV_PROFILE.As you can see, there is no reference to a
GCOV_PROFILEvariable that is named after the sub-directory for which profiling should be enabled.
I've been digging through the git history, and I would swear I hallucinated the whole thing except I have the code in ktest for driving gcov and I swear it used to work :)
Anyways - any thoughts on how we might implement this? I really need a way to specify directories to enable gcov for _without_ monkey patching; that's not a viable workflow in an automated setup.
It seems like if we can get a list of directory prefixes for a path (e.g. given fs/bcachefs/btree_iter.o it would return fs, fs/bcachefs) it should be possible to extend the code you referenced.
Quoted 36 lines · Kent Overstreet
Quoted 22 lines · Peter Oberparleiter
I'm unaware of any kbuild support for setting
GCOV_PROFILEfor a specific sub-directory from the command line, only from within the associatedMakefile. I'm not sure how this could have worked in the past with the provide sample command line.Here's how
GCOV_PROFILEevaluation works (fromscripts/Makefile.lib):ifeq ($(CONFIG_GCOV_KERNEL),y)
_c_flags+= $(if $(patsubst n%,, \
$(GCOV_PROFILE_$(basetarget).o)$(GCOV_PROFILE)$(CONFIG_GCOV_PROFILE_ALL)),\
$(CFLAGS_GCOV))
endifThis bit of
Makefilecode determines whether to add the flags needed to enabled gcov profiling (CFLAGS_GCOV) to the compiler flags for the current compilation unit (_c_flags) by looking at the concatenation of the following variables:- GCOV_PROFILE_<target base name>.o
-GCOV_PROFILE
-CONFIG_GCOV_PROFILE_ALLQuoted 8 lines
gcov flags are only added if this concatenation does not start with an "n", and at least one of these variables is set to a non-empty value other than "n" ("y" typically). The "starts with" part is required to enable precedence for the more specific variable, e.g. an "n" in GCOV_PROFILE_filename.o overwrites a "y" in
GCOV_PROFILE.As you can see, there is no reference to a
GCOV_PROFILEvariable that is named after the sub-directory for which profiling should be enabled.I've been digging through the git history, and I would swear I hallucinated the whole thing except I have the code in ktest for driving gcov and I swear it used to work :)
Quoted 7 lines
Anyways - any thoughts on how we might implement this? I really need a way to specify directories to enable gcov for _without_ monkey patching; that's not a viable workflow in an automated setup.
It seems like if we can get a list of directory prefixes for a path (e.g. given
fs/bcachefs/btree_iter.oit would return fs, fs/bcachefs) it should be possible to extend the code you referenced.
I'll likely not be able to implement this myself, but if you or anyone else wants to go that route, here are my thoughts: $(src) should have the relative source code path that is needed, so here's what needs to be done:
1. Determine how to handle non-letter/digit/underscore characters in the variable name:
a) GCOV_PROFILE_fs/bcachefs => supported by GNU make [1], though discouraged due to possible side-effects
b) GCOV_PROFILE_fs_bcachefs => might cause overlays, e.g. a/b/c and a/b_c both have the same a_b_c suffix
Personally I'd prefer option b)
2. Define a new Makefile variable that contains $(src) with required character replacements (scripts/Kbuild.include might be a good place)
3. Add $(GCOV_PROFILE_$(name_of_that_new_var)) to the code quoted above (scripts/Makefile.lib)
4. Document the use of this new Makefile variable in kernel/gcov/Kconfig and Documentation/dev-tools/gcov.rst
Since this new path-suffix version would be the first GCOV_PROFILE-variable that is primarily intended to be specified on the make command line and not added to a Makefile, it should probably take precedence over all other versions. To achieve that it would need to be specified first in the $(patsubst) statement.
Once implemented, one might also consider extending this new support to other variables like KASAN_SANITIZE or KCOV_INSTRUMENT, since all of these are implemented the same way. I don't know whether that's useful in all cases though.
[1] https://www.gnu.org/software/make/manual/make.html#Using-Variables
Signature · 2 lines
Peter Oberparleiter Linux on IBM Z Development - IBM Germany R&D
Quoted 81 lines · Peter Oberparleiter
Quoted 36 lines · Kent Overstreet
Quoted 22 lines · Peter Oberparleiter
I'm unaware of any kbuild support for setting
GCOV_PROFILEfor a specific sub-directory from the command line, only from within the associatedMakefile. I'm not sure how this could have worked in the past with the provide sample command line.Here's how
GCOV_PROFILEevaluation works (fromscripts/Makefile.lib):ifeq ($(CONFIG_GCOV_KERNEL),y)
_c_flags+= $(if $(patsubst n%,, \
$(GCOV_PROFILE_$(basetarget).o)$(GCOV_PROFILE)$(CONFIG_GCOV_PROFILE_ALL)),\
$(CFLAGS_GCOV))
endifThis bit of
Makefilecode determines whether to add the flags needed to enabled gcov profiling (CFLAGS_GCOV) to the compiler flags for the current compilation unit (_c_flags) by looking at the concatenation of the following variables:- GCOV_PROFILE_<target base name>.o
-GCOV_PROFILE
-CONFIG_GCOV_PROFILE_ALLQuoted 8 lines
gcov flags are only added if this concatenation does not start with an "n", and at least one of these variables is set to a non-empty value other than "n" ("y" typically). The "starts with" part is required to enable precedence for the more specific variable, e.g. an "n" in GCOV_PROFILE_filename.o overwrites a "y" in
GCOV_PROFILE.As you can see, there is no reference to a
GCOV_PROFILEvariable that is named after the sub-directory for which profiling should be enabled.I've been digging through the git history, and I would swear I hallucinated the whole thing except I have the code in ktest for driving gcov and I swear it used to work :)
Quoted 7 lines
Anyways - any thoughts on how we might implement this? I really need a way to specify directories to enable gcov for _without_ monkey patching; that's not a viable workflow in an automated setup.
It seems like if we can get a list of directory prefixes for a path (e.g. given
fs/bcachefs/btree_iter.oit would return fs, fs/bcachefs) it should be possible to extend the code you referenced.I'll likely not be able to implement this myself, but if you or anyone else wants to go that route, here are my thoughts: $(src) should have the relative source code path that is needed, so here's what needs to be done:
1. Determine how to handle non-letter/digit/underscore characters in the variable name:
a) GCOV_PROFILE_fs/bcachefs => supported by GNU make [1], though discouraged due to possible side-effects
b)GCOV_PROFILE_fs_bcachefs=> might cause overlays, e.g. a/b/c and a/b_c both have the samea_b_csuffixPersonally I'd prefer option b)
2. Define a new
Makefilevariable that contains $(src) with required character replacements (scripts/Kbuild.includemight be a good place)3. Add $(GCOV_PROFILE_$(name_of_that_new_var)) to the code quoted above (
scripts/Makefile.lib)4. Document the use of this new
Makefilevariable inkernel/gcov/KconfigandDocumentation/dev-tools/gcov.rstSince this new path-suffix version would be the first GCOV_PROFILE-variable that is primarily intended to be specified on the make command line and not added to a
Makefile, it should probably take precedence over all other versions. To achieve that it would need to be specified first in the $(patsubst) statement.Once implemented, one might also consider extending this new support to other variables like
KASAN_SANITIZEorKCOV_INSTRUMENT, since all of these are implemented the same way. I don't know whether that's useful in all cases though.
Quoted 6 lines
[1] https://www.gnu.org/software/make/manual/make.html#Using-Variables
--
Peter Oberparleiter
Linux on IBM Z Development - IBM Germany R&D
I do not think it is a sensible proposal.
Another idea could be something like CONFIG_GCOV_PROFILE_PATH=fs/bcachefs
or isn't it possible to filter dynamically?
I think ftrace can change filtering dynamically.
I do not know if it is possible for GCOV because I do not use it.
But, the best would be to not implement it at all until we know that it is a wide demand.
The upstream is not a place to cater to every feature request.
Signature · 2 lines
Best Regards Masahiro Yamada
Quoted 91 lines · Masahiro Yamada
Quoted 81 lines · Peter Oberparleiter
Quoted 36 lines · Kent Overstreet
Quoted 22 lines · Peter Oberparleiter
I'm unaware of any kbuild support for setting
GCOV_PROFILEfor a specific sub-directory from the command line, only from within the associatedMakefile. I'm not sure how this could have worked in the past with the provide sample command line.Here's how
GCOV_PROFILEevaluation works (fromscripts/Makefile.lib):ifeq ($(CONFIG_GCOV_KERNEL),y)
_c_flags+= $(if $(patsubst n%,, \
$(GCOV_PROFILE_$(basetarget).o)$(GCOV_PROFILE)$(CONFIG_GCOV_PROFILE_ALL)),\
$(CFLAGS_GCOV))
endifThis bit of
Makefilecode determines whether to add the flags needed to enabled gcov profiling (CFLAGS_GCOV) to the compiler flags for the current compilation unit (_c_flags) by looking at the concatenation of the following variables:- GCOV_PROFILE_<target base name>.o
-GCOV_PROFILE
-CONFIG_GCOV_PROFILE_ALLQuoted 8 lines
gcov flags are only added if this concatenation does not start with an "n", and at least one of these variables is set to a non-empty value other than "n" ("y" typically). The "starts with" part is required to enable precedence for the more specific variable, e.g. an "n" in GCOV_PROFILE_filename.o overwrites a "y" in
GCOV_PROFILE.As you can see, there is no reference to a
GCOV_PROFILEvariable that is named after the sub-directory for which profiling should be enabled.I've been digging through the git history, and I would swear I hallucinated the whole thing except I have the code in ktest for driving gcov and I swear it used to work :)
Quoted 7 lines
Anyways - any thoughts on how we might implement this? I really need a way to specify directories to enable gcov for _without_ monkey patching; that's not a viable workflow in an automated setup.
It seems like if we can get a list of directory prefixes for a path (e.g. given
fs/bcachefs/btree_iter.oit would return fs, fs/bcachefs) it should be possible to extend the code you referenced.I'll likely not be able to implement this myself, but if you or anyone else wants to go that route, here are my thoughts: $(src) should have the relative source code path that is needed, so here's what needs to be done:
1. Determine how to handle non-letter/digit/underscore characters in the variable name:
a) GCOV_PROFILE_fs/bcachefs => supported by GNU make [1], though discouraged due to possible side-effects
b)GCOV_PROFILE_fs_bcachefs=> might cause overlays, e.g. a/b/c and a/b_c both have the samea_b_csuffixPersonally I'd prefer option b)
2. Define a new
Makefilevariable that contains $(src) with required character replacements (scripts/Kbuild.includemight be a good place)3. Add $(GCOV_PROFILE_$(name_of_that_new_var)) to the code quoted above (
scripts/Makefile.lib)4. Document the use of this new
Makefilevariable inkernel/gcov/KconfigandDocumentation/dev-tools/gcov.rstSince this new path-suffix version would be the first GCOV_PROFILE-variable that is primarily intended to be specified on the make command line and not added to a
Makefile, it should probably take precedence over all other versions. To achieve that it would need to be specified first in the $(patsubst) statement.Once implemented, one might also consider extending this new support to other variables like
KASAN_SANITIZEorKCOV_INSTRUMENT, since all of these are implemented the same way. I don't know whether that's useful in all cases though.Quoted 6 lines
[1] https://www.gnu.org/software/make/manual/make.html#Using-Variables
--
Peter Oberparleiter
Linux on IBM Z Development - IBM Germany R&D
Quoted 4 lines
I do not think it is a sensible proposal.
Another idea could be something like
CONFIG_GCOV_PROFILE_PATH=fs/bcachefs
With that it would only be possible to enable profiling for a single subdirectory, and that's not going to be enough; when doing code coverage analysis of a module we may want to profile library code/other modules it depends on.
Quoted 3 lines
or isn't it possible to filter dynamically?
I think ftrace can change filtering dynamically.
I do not know if it is possible for GCOV because I do not use it.
gcov requires different build flags, so no, not possible.
Quoted 72 lines · Masahiro Yamada
Quoted 70 lines · Peter Oberparleiter
Quoted 36 lines · Kent Overstreet
Quoted 22 lines · Peter Oberparleiter
I'm unaware of any kbuild support for setting
GCOV_PROFILEfor a specific sub-directory from the command line, only from within the associatedMakefile. I'm not sure how this could have worked in the past with the provide sample command line.Here's how
GCOV_PROFILEevaluation works (fromscripts/Makefile.lib):ifeq ($(CONFIG_GCOV_KERNEL),y)
_c_flags+= $(if $(patsubst n%,, \
$(GCOV_PROFILE_$(basetarget).o)$(GCOV_PROFILE)$(CONFIG_GCOV_PROFILE_ALL)),\
$(CFLAGS_GCOV))
endifThis bit of
Makefilecode determines whether to add the flags needed to enabled gcov profiling (CFLAGS_GCOV) to the compiler flags for the current compilation unit (_c_flags) by looking at the concatenation of the following variables:- GCOV_PROFILE_<target base name>.o
-GCOV_PROFILE
-CONFIG_GCOV_PROFILE_ALLQuoted 8 lines
gcov flags are only added if this concatenation does not start with an "n", and at least one of these variables is set to a non-empty value other than "n" ("y" typically). The "starts with" part is required to enable precedence for the more specific variable, e.g. an "n" in GCOV_PROFILE_filename.o overwrites a "y" in
GCOV_PROFILE.As you can see, there is no reference to a
GCOV_PROFILEvariable that is named after the sub-directory for which profiling should be enabled.I've been digging through the git history, and I would swear I hallucinated the whole thing except I have the code in ktest for driving gcov and I swear it used to work :)
Quoted 7 lines
Anyways - any thoughts on how we might implement this? I really need a way to specify directories to enable gcov for _without_ monkey patching; that's not a viable workflow in an automated setup.
It seems like if we can get a list of directory prefixes for a path (e.g. given
fs/bcachefs/btree_iter.oit would return fs, fs/bcachefs) it should be possible to extend the code you referenced.I'll likely not be able to implement this myself, but if you or anyone else wants to go that route, here are my thoughts: $(src) should have the relative source code path that is needed, so here's what needs to be done:
1. Determine how to handle non-letter/digit/underscore characters in the variable name:
a) GCOV_PROFILE_fs/bcachefs => supported by GNU make [1], though discouraged due to possible side-effects
b)GCOV_PROFILE_fs_bcachefs=> might cause overlays, e.g. a/b/c and a/b_c both have the samea_b_csuffixPersonally I'd prefer option b)
2. Define a new
Makefilevariable that contains $(src) with required character replacements (scripts/Kbuild.includemight be a good place)3. Add $(GCOV_PROFILE_$(name_of_that_new_var)) to the code quoted above (
scripts/Makefile.lib)4. Document the use of this new
Makefilevariable inkernel/gcov/KconfigandDocumentation/dev-tools/gcov.rst
[...]
Quoted 1 line
I do not think it is a sensible proposal.
Does your objection apply to the proposed implementation, or the feature request itself? If the former, what aspect of the approach do you consider problematic?
Quoted 3 lines
Another idea could be something like
CONFIG_GCOV_PROFILE_PATH=fs/bcachefs
or isn't it possible to filter dynamically?
Using CONFIG symbols would be another way to specify the input list of directories that should be instrumented for GCOV profiling. The output would still need to be a different c_flags value for source files in these directories though. I would assume that getting this information from CONFIG values into c_flags would be a more intrusive change compared to the route via a Makefile variable.
Quoted 2 lines
I think ftrace can change filtering dynamically.
I do not know if it is possible for GCOV because I do not use it.
GCOV instrumentation is a per-source file build-time decision. There's no way to dynamically turn it on or off during run-time.
Quoted 3 lines
But, the best would be to not implement it at all until we know that it is a wide demand.
The upstream is not a place to cater to every feature request.
I don't feel very strongly about this specific feature - it seemed to me that it could be implemented in a non-intrusive way and it provides value to at least one developer.
Signature · 2 lines
Peter Oberparleiter Linux on IBM Z Development - IBM Germany R&D
Quoted 3 lines · Peter Oberparleiter
I don't feel very strongly about this specific feature - it seemed to me that it could be implemented in a non-intrusive way and it provides value to at least one developer.
And it'll get us better tooling integration.
I'm the author of ktest [1], which does interactive and automated kernel testing. With this feature I'll be able to generate gcov/lcov output with a single argument to build-test-kernel, and hopefully with a bit of extra work get code coverage analysis as part of CI runs.
Quoted 8 lines · Peter Oberparleiter
I'll likely not be able to implement this myself, but if you or anyone else wants to go that route, here are my thoughts: $(src) should have the relative source code path that is needed, so here's what needs to be done:
1. Determine how to handle non-letter/digit/underscore characters in the variable name:
Quoted 6 lines
a) GCOV_PROFILE_fs/bcachefs => supported by GNU make [1], though discouraged due to possible side-effects
b)GCOV_PROFILE_fs_bcachefs=> might cause overlays, e.g. a/b/c and a/b_c both have the samea_b_csuffixPersonally I'd prefer option b)
Agreed, feels more consistent
Quoted 5 lines
2. Define a new
Makefilevariable that contains $(src) with required character replacements (scripts/Kbuild.includemight be a good place)3. Add $(GCOV_PROFILE_$(name_of_that_new_var)) to the code quoted above (
scripts/Makefile.lib)
So this is where I was getting stuck, because we really want this to apply to subdirectories (e.g. GCOV_PROFILE_fs_xfs should really also apply to fs/xfs/libxfs).
Do we have existing code for generating a list of path prefixes for a given path?
Quoted 24 lines · Kent Overstreet
Quoted 8 lines · Peter Oberparleiter
I'll likely not be able to implement this myself, but if you or anyone else wants to go that route, here are my thoughts: $(src) should have the relative source code path that is needed, so here's what needs to be done:
1. Determine how to handle non-letter/digit/underscore characters in the variable name:
Quoted 6 lines
a) GCOV_PROFILE_fs/bcachefs => supported by GNU make [1], though discouraged due to possible side-effects
b)GCOV_PROFILE_fs_bcachefs=> might cause overlays, e.g. a/b/c and a/b_c both have the samea_b_csuffixPersonally I'd prefer option b)
Agreed, feels more consistent
Quoted 5 lines
2. Define a new
Makefilevariable that contains $(src) with required character replacements (scripts/Kbuild.includemight be a good place)3. Add $(GCOV_PROFILE_$(name_of_that_new_var)) to the code quoted above (
scripts/Makefile.lib)
Quoted 6 lines
So this is where I was getting stuck, because we really want this to apply to subdirectories (e.g.
GCOV_PROFILE_fs_xfsshould really also apply tofs/xfs/libxfs).Do we have existing code for generating a list of path prefixes for a given path?
Not that I know of. Here's how it could be made to work using Makefile magic alone (as a pure programming exercise :)
This will expand a directory to a list of all parent directories:
# expand_parents(a/b/c) = a/b/c a/b aexpand_parents2 = $(if $(subst .,,$(1)),$(call expand_parents,$(1)),) expand_parents = $(1) $(call expand_parents2,$(patsubst %/,%,$(dir $(1))))
This list could then be turned into variable suffixes:
# flatten_dirs(a/b/c) = a_b_c a_b a flatten_dirs = $(subst /,_,$(call expand_parents,$(1)))
And finally the resulting list of suffixed variables could be evaluated:
# eval_vars(X_,a/b/c) = $(X_a_b_c) $(X_a_b) $(X_a) eval_vars = $(foreach var,$(call flatten_dirs,$(2)),$($(1)$(var)))
So a call like this
$(call eval_vars,GCOV_PROFILE_,a/b/c)
would evaluate to the concatenation of the contents of the following variables:
GCOV_PROFILE_a_b_cGCOV_PROFILE_a_bGCOV_PROFILE_a
The first non-empty variable would then determine whether profiling should be enabled for the associated source file or not. This would even implement the correct order of precedence (specific to generic)
Not sure if this amount of magic is suitable for kbuild though. An alternative, less complex approach would be to move this decision logic to a helper script.
Signature · 2 lines
Peter Oberparleiter Linux on IBM Z Development - IBM Germany R&D
Quoted 14 lines · Peter Oberparleiter
Not that I know of. Here's how it could be made to work using
Makefilemagic alone (as a pure programming exercise :)This will expand a directory to a list of all parent directories:
#
expand_parents(a/b/c)= a/b/c a/b aexpand_parents2= $(if $(subst .,,$(1)),$(call expand_parents,$(1)),)expand_parents= $(1) $(call expand_parents2,$(patsubst %/,%,$(dir $(1))))This list could then be turned into variable suffixes:
#
flatten_dirs(a/b/c)=a_b_ca_baflatten_dirs= $(subst /,_,$(call expand_parents,$(1)))
Quoted 8 lines
And finally the resulting list of suffixed variables could be evaluated:
#
eval_vars(X_,a/b/c)= $(X_a_b_c) $(X_a_b) $(X_a)eval_vars= $(foreach var,$(call flatten_dirs,$(2)),$($(1)$(var)))So a call like this
$(call eval_vars,GCOV_PROFILE_,a/b/c)
I just hooked it up and it works perfectly - I owe you a beer :)
Quoted 7 lines
would evaluate to the concatenation of the contents of the following variables:
GCOV_PROFILE_a_b_cGCOV_PROFILE_a_bGCOV_PROFILE_a
Quoted 7 lines
The first non-empty variable would then determine whether profiling should be enabled for the associated source file or not. This would even implement the correct order of precedence (specific to generic)
Not sure if this amount of magic is suitable for kbuild though. An alternative, less complex approach would be to move this decision logic to a helper script.
Now that I've spent a couple days starting to wrap my head more around make, it doesn't look terribly magic to me. I'd hate to have to spawn off a subshell for this.
Quoted 17 lines · Kent Overstreet
Quoted 16 lines · Peter Oberparleiter
Not that I know of. Here's how it could be made to work using
Makefilemagic alone (as a pure programming exercise :)This will expand a directory to a list of all parent directories:
#
expand_parents(a/b/c)= a/b/c a/b aexpand_parents2= $(if $(subst .,,$(1)),$(call expand_parents,$(1)),)expand_parents= $(1) $(call expand_parents2,$(patsubst %/,%,$(dir $(1))))This list could then be turned into variable suffixes:
#
flatten_dirs(a/b/c)=a_b_ca_baflatten_dirs= $(subst /,_,$(call expand_parents,$(1)))And finally the resulting list of suffixed variables could be evaluated:
Quoted 8 lines
Quoted 6 lines
#
eval_vars(X_,a/b/c)= $(X_a_b_c) $(X_a_b) $(X_a)eval_vars= $(foreach var,$(call flatten_dirs,$(2)),$($(1)$(var)))So a call like this
$(call eval_vars,GCOV_PROFILE_,a/b/c)
I just hooked it up and it works perfectly - I owe you a beer :)
Glad to be of help!
Quoted 11 lines
Quoted 11 lines
would evaluate to the concatenation of the contents of the following variables:
GCOV_PROFILE_a_b_cGCOV_PROFILE_a_bGCOV_PROFILE_aThe first non-empty variable would then determine whether profiling should be enabled for the associated source file or not. This would even implement the correct order of precedence (specific to generic)
Quoted 7 lines
Quoted 3 lines
Not sure if this amount of magic is suitable for kbuild though. An alternative, less complex approach would be to move this decision logic to a helper script.
Now that I've spent a couple days starting to wrap my head more around make, it doesn't look terribly magic to me. I'd hate to have to spawn off a subshell for this.
Leaving the method of how this input data is processed inside kbuild aside for a moment, I'm wondering if specifying the list of directories via a CONFIG symbol instead of make parameters would work equally well.
Initially I thought this would be too complex to process using Makefile functions alone, but with the logic shown above this could be relatively simple to achieve. Also having given this some more thought, a CONFIG symbol indeed seems like a better fit considering aspects such as reproducibility of builds and config symbol documentation.
The CONFIG symbols could look something like:
- CONFIG_GCOV_PROFILE_INCLUDE: list of directories to include in profiling
- CONFIG_GCOV_PROFILE_EXCLUDE: list of directories to exclude from profiling
where the precedence would be: exclude list > include list > GCOV_PROFILE_* specified in Makefiles > CONFIG_GCOV_PROFILE_ALL
Sub-directory handling could work similar to how you described it for the make parameter, i.e. an include/exclude statement for a parent directory would also apply to sub-directories.
What this approach work for your use case? Note that I'm not asking you to implement this - I just want to get a better picture of how a generic solution could look like.
Signature · 2 lines
Peter Oberparleiter Linux on IBM Z Development - IBM Germany R&D
Quoted 20 lines · Peter Oberparleiter
Leaving the method of how this input data is processed inside kbuild aside for a moment, I'm wondering if specifying the list of directories via a CONFIG symbol instead of make parameters would work equally well.
Initially I thought this would be too complex to process using
Makefilefunctions alone, but with the logic shown above this could be relatively simple to achieve. Also having given this some more thought, a CONFIG symbol indeed seems like a better fit considering aspects such as reproducibility of builds and config symbol documentation.The CONFIG symbols could look something like:
-
CONFIG_GCOV_PROFILE_INCLUDE: list of directories to include in profiling
-CONFIG_GCOV_PROFILE_EXCLUDE: list of directories to exclude from profilingwhere the precedence would be: exclude list > include list > GCOV_PROFILE_* specified in Makefiles >
CONFIG_GCOV_PROFILE_ALL
Quoted 7 lines
Sub-directory handling could work similar to how you described it for the make parameter, i.e. an
include/excludestatement for a parent directory would also apply to sub-directories.What this approach work for your use case? Note that I'm not asking you to implement this - I just want to get a better picture of how a generic solution could look like.
Yeah, that would work fine. I'm assuming the config option list would just be a comma separated string?
btw, here's the results - fully automated code coverage analysis integrated into the CI:
https://evilpiepirate.org/~testdashboard/ci?branch=bcachefs&commit=eac40840098bfbb5fa1711f6bbce71b27bbccb89
https://evilpiepirate.org/~testdashboard/c/eac40840098bfbb5fa1711f6bbce71b27bbccb89/lcov/
And the result of the makefile work is that adding a gcov variant of an existing test is as easy as:
https://evilpiepirate.org/git/ktest.git/tree/tests/bcachefs/gcov-xfstests.ktest