Makefile wildcard (static rule?) with phony - makefile

I'm just starting to really grok the inner workings of make. Yet I do not understand why the following doesn't work:
test%: test%.foo
#echo $#
#echo $<
all: test1 test2
.PHONY: all test1 test2
Expected behavior:
$ make
test1
test1.foo
test2
test2.foo
# 1,2 Order not important
However, I get:
$ make
make: Nothing to be done for `all'.
("make all", "make test1", etc make no difference).
Can someone explain why the PHONY test rules aren't being executed?

Excerpt from the GNU make manual.
Since it knows that phony targets do not name actual files that could
be remade from other files, make skips the implicit rule search for
phony targets (see section Using Implicit Rules). This is why
declaring a target phony is good for performance, even if you are not
worried about the actual file existing.
This means that as your test1 and test2 targets are phony, make does not search for implicit rules for them. Even if what you use is more accurately named pattern rules, all pattern rules are implicit rules.

Related

Makefile: Match-Anything Pattern Rule

I have a main Makefile that calls Makefiles placed in subfolders. For testing purposes I would like to add Match-Anything rule at the end of main Makefile.
This rule would be:
%:
make -e -C subdir $#
Are there any contrarguments for such a rule?
I think you meant "con", as in "pro or con", not "cont". "Con" is short for the Latin "contra".
For your question, the downside of adding a new "match anything" rule is that any file that doesn't exist, will try to be created using this rule. For example suppose you run include foo.mk and foo.mk doesn't exist... make will attempt to build foo.mk by running your rule. Basically it can be confusing. It's possible there could be some performance impact; if you run make -d and examine it you should see if your match-anything rule is being used at all during a normal build. I'm not sure if there are any other serious downsides.
By the way you should always only use the make variable $(MAKE) when invoking a sub-make in a recipe; never use the raw make command.

defer prerequisite expansion until after a (different) target creation

I want to be able use the result of a target created in a rule in the prerequisite of another rule in GNU make. So for example:
PREREQ = $(shell echo "reading target1" >&2; cat target1)
target1:
echo "prereq" > $#
target2: target1 $(PREREQ)
echo foo > $#
target2 should depend on prereq as read from the target1 file, but that is not in the file until the target1 recipe is executed.
Granted this is very contrived example with I am sure lots of suggestions about how to refactor this particular example but I'm not looking to refactor this example. This is just a simplified example of my more complicated problem where I need to derive prerequisites from the contents of a file that is not created until a recipe in the Makefile is executed.
The question is, [how] can I make expansion of $(PREREQ) (and therefore the execution of the $(shell cat target1) defer until after the target1 rule is actually executed?
Update: I tried .SECONDARYEXPANSION: but that doesn't seem to do the job:
$ make -d target2
...
reading target1
cat: target1: No such file or directory
...
Updating goal targets....
Considering target file 'target2'.
File 'target2' does not exist.
Considering target file 'target1'.
File 'target1' does not exist.
Finished prerequisites of target file 'target1'.
Must remake target 'target1'.
echo "prereq" > target1
[ child management ]
Successfully remade target file 'target1'.
Finished prerequisites of target file 'target2'.
Must remake target 'target2'.
echo foo > target2
[ child management ]
Successfully remade target file 'target2'.
As you can see, "reading target" was only printed once at the very beginning demonstrating that PREREQ is not expanded again due to the .SECONDEXPANSION: and the list of targets Considered for target2 did not include prereq.
Deferring the expansion of the prerequisite $(PREREQ) can be achieved by conditionally creating the target2 and relying on recursion:
ifndef expand-prereq
target2: target1
$(MAKE) --no-print-directory -f $(lastword $(MAKEFILE_LIST)) $# expand-prereq=y
else
target2: target1 $(PREREQ)
echo foo > $#
endif
The first time make runs for this makefile, the variable expand-prereq is not defined and therefore, the first targe2 rule is generated as a result of the conditional. This kind of dummy rule makes possible to update target1 without expanding $(PREREQ).
Matching this rule results in target1 being updated (since target1 is a prerequisite of this rule) and make being called recursively for the same makefile and with target2 as target.
The second time make is (recursively) invoked, the variable expand-prereq was defined by means of the command-line argument expand-prereq=y, so the second target2 rule is generated as a result of the else branch this time. This rule is the one that actually produces the target target2. Note that before this rule can be matched, target1 has been already created as a side effect of the first dummy rule, so the expansion of $(PREREQ) happens after target1 has been created (what you were looking for).
You could write the complete rule for target2 to a separate file and -include it:
Including Other Makefiles
How Makefiles Are Remade
The exact mechanics will depend on your specific use case, and it may well be impossible to achieve what we need using this approach, but it supports a variety of styles for automated dependency generation.
There are several solutions:
GNU make 4.4 has been released! Haven't tried yet, but the release notes claim that secondary expansion only expands the prerequisites when they're considered. Furthermore you can delay execution with the .WAIT special prerequisite. That works fine, I tested. If .WAIT really delays the second expansion of the prerequisites after .WAIT, you're good.
Recursive make. Restart make after the prerequisites for the second rule were updated. This is a bit lame solution, can't recommend it.
Produce the prerequisites into make include file(s). Make automatically restarts after updating include files (re-exec). I'm currently using this method, and it works great. Better than recursive make but still slow, as all the makefiles have to be parsed again. A possible solution is comparing the old and new prerequisites, and only updating the include file, if its content changed. The second rule also needs to be modified to do nothing, if the content changed. Make will run all rules before restarting, but if they don't update their targets, after the restart they'll be executed again, now with the proper prerequisites. An interesting feature of make is that you can define make variables inside a recipe, and use them in other recipes (but not in the prereq list, unless #1 above reall works).

a surprising (?) behaviour of GNU Make when using ``%`` as target

Consider the following Makefile
foo:
#echo '$#'
test:
#echo '$#'
#echo '---'
# Catch-all target
%: test
#echo '+++'
#echo '$#'
When issuing make bar the following is the console output:
$ make bar
test
---
+++
Makefile
+++
bar
I would like to understand the origin of Makefile which shows it is received as argument at some point, and also how to get rid of it in such a scheme. This is using
GNU Make 4.1
Built for x86_64-apple-darwin13.4.0
GNU make treats the makefile itself as a target that needs to be updated. See How Makefiles Are Remade:
... after reading in all makefiles, make will consider each as a goal target and attempt to update it. If a makefile has a rule which says how to update it (found either in that very makefile or in another one) or if an implicit rule applies to it (see Using Implicit Rules), it will be updated if necessary...
If you know that one or more of your makefiles cannot be remade and you want to keep make from performing an implicit rule search on them, perhaps for efficiency reasons, you can use any normal method of preventing implicit rule look-up to do so. For example, you can write an explicit rule with the makefile as the target, and an empty recipe (see Using Empty Recipes).
Hence, the catch-all-target % is used to update Makefile.
Makefiles often do not have to be updated, so it is customary to add an empty rule for that:
Makefile : ;

error using makefile, targets and %

I'm trying to debug the following code:
TESTS=$(shell cat yoursourcefile)
all: $(TESTS)
%: compile_design
compile $#_tb.vhd >> log_file.log
simulate $#
I got this error:
makefile_tb.vhd >> log_file.log
as if makefile is a target
this error disappears when I add a character or more before %:
T%: compile_design
compile $#_tb.vhd >> log_file.log
simulate $#
This works but implies that all my targets starts with "T" which is not always the case.
My questions are:
what's exactly the function of % here ?
How to get rid of this error?
As suggested, I added
makefile: ; $#:
at the end, so I have now:
TESTS=$(shell cat yoursourcefile)
all: $(TESTS)
%: compile_design
compile $#_tb.vhd >> log_file.log
simulate $#
makefile: ; $#:
then when I do:
make all
I get [all] error2 all_tb.vhd >> log_file.log
but all_tb.vhd does not exist !
The %: compile_design rule is a "match-anything" pattern rule. It says "hey make, if you ever want to build any file, with any name, then you can do it by running these commands. Oh and by the way, if you have a file you want to build and it's older than the compile_design file, then you need to rebuild it". Generally you want to avoid match-anything rules, but if your target names truly have no specific pattern, you can't.
When you add the T before it then it tells make that instead of any file, that rule can only build files that begin with T.
The reason make is trying to rebuild the makefile is that GNU make has a special feature that allows it to remake its own makefiles. So after it reads its makefile it will try to re-make it. Normally this has no effect because there's no rule to build a makefile, but here you've added a rule that you've told make can build anything. Adding the T keeps the pattern from matching Makefile because Makefile doesn't begin with T.
The simplest thing for you to do is define an explicit rule for the makefile: make always chooses an explicit rule, if it exists, over an implicit rule like a pattern rule:
Makefile: ; #:
This creates an explicit rule that does nothing (: is the shell built-in command that does nothing).

Ensuring a prerequisite exists for a pattern rule

I understand that an "explicit" pattern rule will take precedence on its implicit counterpart when its prerequisites can be made.
all: src/foo.o
src/%.o: makefile my_haeder.h src/%.c
echo Do something with those source files
If there is a typo for "my_header.h", the implicit rule for %.o will take precedence. Not only my recipe will not be executed, but touching the prerequisites will not trigger the rule. Actually it is the second point which is of interest for me.
The make documentation offers a verification using static pattern rules:
SET_OF_FILES=src/foo.o
all: src/foo.o
$(SET_OF_FILES): src/%.o: makefile my_haeder.h src/%.c
echo Do something with those source files
This results in:
gmake: *** No rule to make target `src/my_haeder.h', needed by `src/foo.o'. Stop.
Though a larger rule, that solution is nice, as long as I don't have to add a rule for which the stem could overlap:
SET_OF_FILES=src/foo.o src/subsrc/bar.o
all: src/foo.o
$(SET_OF_FILES): src/%.o: makefile my_header.h src/%.c
echo Do something with those source files
$(SET_OF_FILES): src/subsrc/%.o: makefile my_header.h src/subsrc/%.c
echo Do something with those other source files
Which results in:
makefile:8: target `src/foo.o' doesn't match the target pattern
makefile:9: warning: overriding commands for target `src/foo.o'
makefile:6: warning: ignoring old commands for target `src/foo.o'
makefile:9: warning: overriding commands for target `src/subsrc/bar.o'
makefile:6: warning: ignoring old commands for target `src/subsrc/bar.o'
The first message is here because I didn't bother $(filter)ing SET_OF_FILES. I don't know how to solve the next warnings, which for any reviewer would mean "something's wrong".
Is there another (more elegant) way to verify that the prerequisites are actually feasible, in order to avoid dropping the explicit pattern rule?
(using GNU Make 3.79.1 win32)
Add a separate rule to check for your prerequisites
all: prereqs src/foo.o
prereqs: Makefile my_header.h
src/%.o: src/%.c Makefile my_header.h
echo Do something with those source files
src/subsrc/%.o: src/subsrc/%.c Makefile my_header.h
echo Do something with those other source files
which will give you:
make
make: *** No rule to make target 'my_header.h', needed by 'prereqs'. Stop

Resources