Makefile: ifeq directive compares special variable to constant does not work - variables

I have a little issue with my Makefile. I want make to change commands regarding the working directory. I added a conditional directive to the rule testing the current target's directory ($(*D)).
The thing is that make always goes to the second branch of my test, even when my file is in mySpecialDirectory and echo indeed prints "mySpecialDirectory".
.c.o .cpp.o .cc.o:
ifeq ($(*D),mySpecialDirectory)
#echo "I'm in mySpecialDirectory! \o/"
$(CC) $(DEBUG_FLAGS) $(MYSPECIALFLAGS) -c $< -o $#
else
#echo "Failed! I'm in $(*D)"
$(CC) $(DEBUG_FLAGS) $(NOTTHATSPECIALFLAGS) -c $< -o $#
endif

This is the expected behavior.
Conditional Statements
All instances of conditional syntax are parsed immediately, in their entirety; this includes the ifdef, ifeq, ifndef, and ifneq forms. Of course this means that automatic variables cannot be used in conditional statements, as automatic variables are not set until the command script for that rule is invoked. If you need to use automatic variables in a conditional you must use shell conditional syntax, in your command script proper, for these tests, not make conditionals.
$(*D) is an automatic variable.
Instead, consider doing:
.c.o .cpp.o .cc.o:
$(CC) $(DEBUG_FLAGS) $(if $(subst mySpecialDirectory,,$(*D)),$(NOTTHATSPECIALFLAGS),$(MYSPECIALFLAGS)) -c $< -o $#
The idea is to abuse $(subst) into some equality testing by replacing mySpecialDirectory with the empty string. Then, if $(*D) expansion equals to mySpecialDirectory, it is fully replaced by the empty string and the else-part of $(if) gets evaluated as per:
$(if condition,then-part[,else-part])
The if function provides support for conditional expansion in a functional context (as opposed to the GNU make makefile conditionals such as ifeq (see Syntax of Conditionals).
The first argument, condition, first has all preceding and trailing whitespace stripped, then is expanded. If it expands to any non-empty string, then the condition is considered to be true. If it expands to an empty string, the condition is considered to be false.
Notice the flip in this hack between the then-part and the else-part.
Hope this helps!

Related

Compiler option containing doublequotes and space breaks compilation

I need to add a compiler option that contains doublequotes and space to a target in my project. I use target_compile_options to add an option.
When I add the option as I would do it on a command line CMake escapes " with \ and adds extra " (I assume the second is because of whitespace), thus breaking the compilation.
CMakeLists.txt command:
target_compile_options(myTarget -Woption="PARAM VALUE" -otherOption)
Output:
compiler.exe "-Woption=\"PARAM VALUE\"" -otherOption -o myTarget.o -c myTarget.ext
Desired output:
compiler.exe -Woption="PARAM VALUE" -otherOption -o myTarget.o -c myTarget.ext
I assume I have two problems. One is escaping doublequotes and the second one is escaping whitespace.
Question
Is it possible to add an option containing doublequotes and whitespace to compiler options? If so can somebody point me to a right direction on how to do it?
Notes
I use a cross compiler thus I made the example generic but in case it plays a role this is how the target is compiled:
set(CMAKE_${lang}_COMPILE_OBJECT "<CMAKE_${lang}_COMPILER> <FLAGS> <INCLUDES> -o <OBJECT> -c <SOURCE>")
cmake version 3.12
Build platform: Windows

How do I include a literal double quote in a custom CMake command?

I'm trying to create a custom command that runs with some environment variables, such as LDFLAGS, whose value needs to be quoted if it contains spaces:
LDFLAGS="-Lmydir -Lmyotherdir"
I cannot find a way to include this argument in a CMake custom command, due to CMake's escaping rules. Here's what I've tried so far:
COMMAND LDFLAGS="-Ldir -Ldir2" echo blah VERBATIM)
yields "LDFLAGS=\"-Ldir -Ldir2\"" echo blah
COMMAND LDFLAGS=\"-Ldir -Ldir2\" echo blah VERBATIM)
yields LDFLAGS=\"-Ldir -Ldir2\" echo blah
It seems I either get the whole string quoted, or the escaped quotes don't resolve when used as part of the command.
I would appreciate either a way to include the literal double-quote or as an alternative a better way to set environment variables for a command. Please note that I'm still on CMake 2.8, so I don't have the new "env" command available in 3.2.
Note that this is not a duplicate of When to quote variables? as none of those quoting methods work for this particular case.
The obvious choice - often recommended when hitting the boundaries of COMMAND especially with older versions of CMake - is to use an external script.
I just wanted to add some simple COMMAND only variations that do work and won't need a shell, but are - I have to admit - still partly platform dependent.
One example would be to put only the quoted part into a variable:
set(vars_as_string "-Ldir -Ldir2")
add_custom_target(
QuotedEnvVar
COMMAND env LD_FLAGS=${vars_as_string} | grep LD_FLAGS
)
Which actually does escape the space and not the quotes.
Another example would be to add it with escaped quotes as a "launcher" rule:
add_custom_target(
LauncherEnvVar
COMMAND env | grep LD_FLAGS
)
set_target_properties(
LauncherEnvVar
PROPERTIES RULE_LAUNCH_CUSTOM "env LD_FLAGS=\"-Ldir -Ldir2\""
)
Edit: Added examples for multiple quoted arguments without the need of escaping quotes
Another example would be to "hide some of the complexity" in a function and - if you want to add this to all your custom command calls - use the global/directory RULE_LAUNCH_CUSTOM property:
function(set_env)
get_property(_env GLOBAL PROPERTY RULE_LAUNCH_CUSTOM)
if (NOT _env)
set_property(GLOBAL PROPERTY RULE_LAUNCH_CUSTOM "env")
endif()
foreach(_arg IN LISTS ARGN)
set_property(GLOBAL APPEND_STRING PROPERTY RULE_LAUNCH_CUSTOM " ${_arg}")
endforeach()
endfunction(set_env)
set_env(LDFLAGS="-Ldir1 -Ldir2" CFLAGS="-Idira -Idirb")
add_custom_target(
MultipleEnvVar
COMMAND env | grep -E 'LDFLAGS|CFLAGS'
)
Alternative (for CMake >= 3.0)
I think what we actually are looking for here (besides the cmake -E env ...) is named Bracket Argument and does allow any character without the need of adding backslashes:
set_property(
GLOBAL PROPERTY
RULE_LAUNCH_CUSTOM [=[env LDFLAGS="-Ldir1 -Ldir2" CFLAGS="-Idira -Idirb"]=]
)
add_custom_target(
MultipleEnvVarNew
COMMAND env | grep -E 'LDFLAGS|CFLAGS'
)
References
0005145: Set environment variables for ADD_CUSTOM_COMMAND/ADD_CUSTOM_TARGET
How to modify environment variables passed to custom CMake target?
[CMake] How to set environment variable for custom command
cmake: when to quote variables?
You need three backslashes. I needed this recently to get a preprocessor define from PkgConfig and apply it to my C++ flags:
pkg_get_variable(SHADERDIR movit shaderdir)
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -DSHADERDIR=\\\"${SHADERDIR}\\\"")
Florian's answer is wrong on several counts:
Putting the quoted part in a variable makes no difference.
You should definitely use VERBATIM. It fixes platform-specific quoting bugs.
You definitely shouldn't use RULE_LAUNCH_CUSTOM for this. It isn't intended for this and only works with some generators.
You shouldn't use env as the command. It isn't available on Windows.
It turns out the real reason OPs code doesn't work is that CMake always fully quotes the first word after COMMAND because it's supposed to be the name of an executable. You simply shouldn't put environment variables first.
For example:
add_custom_command(
OUTPUT q1.txt
COMMAND ENV_VAR="a b" echo "hello" > q1.txt
VERBATIM
)
add_custom_target(q1 ALL DEPENDS q1.txt)
$ VERBOSE=1 make
...
"ENV_VAR=\"a b\"" echo hello > q1.txt
/bin/sh: ENV_VAR="a b": command not found
So how do you pass an environment variable with spaces? Simple.
add_custom_command(
OUTPUT q1.txt
COMMAND ${CMAKE_COMMAND} -E env ENV_VAR="a b" echo "hello" > q1.txt
VERBATIM
)
Ok, I removed my original answer as the one proposed by #Florian is better. There is one additional tweak needed for multiple quoted args. Consider a list of environment variables as such:
set(my_env_vars LDFLAGS="-Ldir1 -Ldir2" CFLAGS="-Idira -Idirb")
In order to produce the desired expansion, convert to string and then replace ; with a space.
set(my_env_string "${my_env_vars}") #produces LDFLAGS="...";CFLAGS="..."
string(REPLACE ";" " " my_env_string "${my_env_string}")
Then you can proceed with #Florian's brilliant answer and add the custom launch rule. If you need semicolons in your string then you'll need to convert them to something else first.
Note that in this case I didn't need to launch with env:
set_target_properties(mytarget PROPERTIES RULE_LAUNCH_CUSTOM "${my_env_string}")
This of course depends on your shell.
On second thought, my original answer is below as I also have a case where I don't have access to the target name.
set(my_env LDFLAGS=\"-Ldir -Ldir2" CFLAGS=\"-Idira -Idirb\")
add_custom_command(COMMAND sh -c "${my_env} grep LDFLAGS" VERBATIM)
This technique still requires that the semicolons from the list->string conversion be replaced.
Some folks suggest to use ${CMAKE_COMMAND} and pass your executable as an argument, e.g:
COMMAND ${CMAKE_COMMAND} -E env "$(WindowsSdkDir)/bin/x64/makecert.exe" ...
That worked for me.

How do I escape a "$" in bitbake/yocto?

One of my recipes in Yocto need to create a file containing a very specific line, something like:
${libdir}/something
To do this, I have the recipe task:
do_install() {
echo '${libdir}/something' >/path/to/my/file
}
Keeping in mind that I want that string exactly as shown, I can't figure out how to escape it to prevent bitbake from substituting in its own value of libdir.
I originally thought the echo command with single quotes would do the trick (as it does in the bash shell) but bitbake must be interpreting the line before passing it to the shell. I've also tried escaping it both with $$ and \$ to no avail.
I can find nothing in the bitbake doco about preventing variable expansion, just stuff to do with immediate, deferred and Python expansions.
What do I need to do to get that string into the file as is?
Bitbake seems to have particular issues in preventing expansion from taking place. Regardless of whether you use single or double quotes, it appears that the variables will be expanded before being passed to the shell.
Hence, if you want them to not be expanded, you need to effectively hide them from BitBake, and this can be done with something like:
echo -e '\x24{libdir}/something' >/path/to/my/file
This uses the hexadecimal version of $ so that BitBake does not recognise it as a variable to be expanded.
You do need to ensure you're running the correct echo command however. Under some distros (like Ubuntu), it might run the sh-internal echo which does not recognise the -e option. In order to get around that, you may have to run the variant of echo that lives on the file system (and that does recognise that option):
/bin/echo -e '\x24{libdir}/something' >/path/to/my/file
By default this task will be executed as shell function via /bin/sh, but it depends on your system what it will be as you can have a symlink named /bin/sh pointing to bash. The BitBake's manual prevents from using bashism syntax though.
You can consider just adding this task in your recipe as python function:
python do_install () {
with open('/path/to/your/file', 'a') as file:
file.write('${libdir}/something')
}
'a' stands for append.
This should eliminate the problem with variable expansion.
There is no standard way to escape these sorts of expressions that I am aware of, other than to try to break up the expression - accordingly this should work:
do_install() {
echo '$''{libdir}/something' >/path/to/my/file
}
The best solution is simply this:
bitbake_function() {
command $libdir/whatever
}
Bitbake will only expand ${libdir}; $libdir is passed through verbatim.
We don't have to worry about dollar signs that are not followed by {, and in this case, there is no need for libdir to be wrapped in braces.
The only time we run into a problem with just $foo is if we have something like ${foo}bar where the braces are required as delimiters so that bar isn't included into the variable name. In that situation, there are other solutions, such as for instance generating the shell syntax "$foo"bar. This is less cryptic than resorting to \x24.
If you need to use $ in variable assignment, remember that bitbake won't evaluate $whatever so you have to escape it for the underlying shell.
For instance I set gcc/ld Rpath option to use $ORIGIN keyword this way:
TARGET_LDFLAGS_append = " -Wl,-rpath-link=\\$$ORIGIN"
https://lists.yoctoproject.org/pipermail/yocto/2017-September/037820.html
You can define a variable to be a literal dollar sign.
DOLLAR = "$"
do_install() {
echo '${DOLLAR}{libdir}/something' >/path/to/my/file
}
no extra quoting required.

make targets depend on variables

I want (GNU) make to rebuild when variables change. How can I achieve this?
For example,
$ make project
[...]
$ make project
make: `project' is up to date.
...like it should, but then I'd prefer
$ make project IMPORTANTVARIABLE=foobar
make: `project' is up to date.
to rebuild some or all of project.
Make wasn't designed to refer to variable content but Reinier's approach shows us the workaround. Unfortunately, using variable value as a file name is both insecure and error-prone. Hopefully, Unix tools can help us to properly encode the value. So
IMPORTANTVARIABLE = a trouble
# GUARD is a function which calculates md5 sum for its
# argument variable name. Note, that both cut and md5sum are
# members of coreutils package so they should be available on
# nearly all systems.
GUARD = $(1)_GUARD_$(shell echo $($(1)) | md5sum | cut -d ' ' -f 1)
foo: bar $(call GUARD,IMPORTANTVARIABLE)
#echo "Rebuilding foo with $(IMPORTANTVARIABLE)"
#touch $#
$(call GUARD,IMPORTANTVARIABLE):
rm -rf IMPORTANTVARIABLE*
touch $#
Here you virtually depend your target on a special file named $(NAME)_GUARD_$(VALUEMD5) which is safe to refer to and has (almost) 1-to-1 correspondence with variable's value. Note that call and shell are GNU Make extensions.
You could use empty files to record the last value of your variable by using something like this:
someTarget: IMPORTANTVARIABLE.$(IMPORTANTVARIABLE)
#echo Remaking $# because IMPORTANTVARIABLE has changed
touch $#
IMPORTANTVARIABLE.$(IMPORTANTVARIABLE):
#rm -f IMPORTANTVARIABLE.*
touch $#
After your make run, there will be an empty file in your directory whose name starts with IMPORTANTVARIABLE. and has the value of your variable appended. This basically contains the information about what the last value of the variable IMPORTANTVARIABLE was.
You can add more variables with this approach and make it more sophisticated using pattern rules -- but this example gives you the gist of it.
You probably want to use ifdef or ifeq depending on what the final goal is. See the manual here for examples.
I might be late with an answer, but here is another way of doing such a dependency with Make conditional syntax (works on GNU Make 4.1, GNU bash, Bash on Ubuntu on Windows version 4.3.48(1)-release (x86_64-pc-linux-gnu)):
1 ifneq ($(shell cat config.sig 2>/dev/null),prefix $(CONFIG))
2 .PHONY: config.sig
3 config.sig:
4 #(echo 'prefix $(CONFIG)' >config.sig &)
5 endif
In the above sample we track the $(CONFIG) variable, writing it's value down to a signature file, by means of the self-titled target which is generated under condition when the signature file's record value is different with that of $(CONFIG) variable. Please, note the prefix on lines 1 and 4: it is needed to distinct the case, when signature file doesn't exist yet.
Of course, consumer targets specify config.sig as a prerequisite.

Gnu Makefile, compile should fail

I have a collection of examples that I want to make sure they fail to compile. What is the best way to to that with a *GNU Makefile?
test_nocompile: $(NOCOMPILE_CPP)
for cpp in $(NOCOMPILE_CPP) ; do \
echo === $$cpp === ; \
if $(CXX) $(CXXFLAGS) -c -o fail.o $$cpp ; then echo ok ; else exit 1; fi ; \
done
As you can see, I have several difficulties here:
Accessing the shell-for variable cpp: $cpp, \$cpp and $$cpp both do not work.
Even with my if the make stops after the first g++-compile fails. But thats exactly what I want. I want failing to g++-compile to be considered the correct behaviour.
Shell-for-loops in Makefiles are not the best idea. Is there a better way? I can not think of one, because since I expect the compiles to fail, I do not have a correct target, right?
Advanced and optional:
If I could manage to have the fail-check above working, I could try a second pass of compilation, this time with an additional -DEXCLUDE_FAIL, which takes out the offending line from my examples, and then the code should compile. Any idea?
or should write a Python script for that...? ;-)
Edit:
I think my "advanced and optional" gave me a very good idea right now. I could use makes dependency checking again. Just a rough sketch here:
NOCOMPILE_CPP := $(wildcard nocompile/ *.cpp)
NOCOMPILE_XFAILS := $(addsuffix .xfail,$(basename $(NOCOMPILE_CPP)))
%.xfail: %.cpp
if $(CXX) $(CXXFLAGS) -o $# $< ; then exit 1 ; else echo OK ; fi
$(CXX) $(CXXFLAGS) -DEXCLUDE_FAILCODE -o $# $<
test_nocompile: $(NOCOMPILE_XFAILS)
Is this elegant? Then I only have to work out how -DEXCLUDE_FAILCODE can make the failing tests work.... Non-trivial, but doable. I think that could do it, right?
Works for me. What do you get in echo?
You need to negate the condition in if then. Right now it quits when the file doesn't compile and AFAIU you need the opposite.
Why it's not a good idea? But you can write a script that will call $(CXX) and return a proper error code (0 if it doesn't compile). Then you may have normal targets with this script. I'm not very good with specifics of GNU make, probably it's possible with the builtin stuff.
Advanced & optional:
1. Let's first make the thing work:)
2. Personally i don't use python, therefore don't see a need here =:P