List goals/targets in GNU make that contain variables in their definition
I have a fairly large makefile that creates a number of targets on the fly by开发者_Go百科 computing names from variables. (eg foo$(VAR) : $(PREREQS)). Is there any way that gnu make can be convinced to spit out a list of targets after it has expanded these variables?
I'd like to be able to get the targets for an aribitrary makefile. I'm trying to write a completion function for my shell.
make -qp | awk -F':' '/^[a-zA-Z0-9][^$#\/\t=]*:([^=]|$)/ {split($1,A,/ /);for(i in A)print A[i]}'
Taken from the make arg completion, which works like a charm.
Can you parse the output from make -pn
(i.e. make --print-data-base --dry-run
)? It prints out all the variables, rules, implicit rules and which commands will be run in laborious detail.
I'm not sure if this is only a gnu make thing, but this works well:
make help
Several responders have suggested using make -pn
, which will print the database of rules but not execute anything -- more or less. The problem with this approach is that -n
does still invoke all recursive makes, and it does still do a lot more work than necessary, because it prints out every command that it would have invoked in a regular build. A more efficient solution would be to create a trivial makefile, dummy.mk, with these contents:
__all_targets__: ; #no-op
Now invoke make as make -p -f Makefile -f dummy.mk __all_targets__
. On any substantial build, the difference in the amount of output generated by make is significant. For example:
$ gmake -pn | wc
138985 2632330 69612711
$ gmake -f Makefile -f /tmp/dummy.mk -pn __all_targets__ | wc
21673 93437 878985
Execution time was dramatically better as well -- 2.063s for the first version, 0.059s for the second.
Check out bash completion for make on GitHub.
Edit: FYI the debian bash completion git repository in another answer now incorporates an enhanced version of this script tailored to bash completion use cases.
#!/bin/bash
SCRIPT='
/^# Make data base/,/^# Files/d # skip until files section
/^# Not a target/,+1 d # following target isnt
/^\.PHONY:/ d # special target
/^\.SUFFIXES:/ d # special target
/^\.DEFAULT:/ d # special target
/^\.PRECIOUS:/ d # special target
/^\.INTERMEDIATE:/ d # special target
/^\.SECONDARY:/ d # special target
/^\.SECONDEXPANSION/ d # special target
/^\.DELETE_ON_ERROR:/ d # special target
/^\.IGNORE:/ d # special target
/^\.LOW_RESOLUTION_TIME:/ d # special target
/^\.SILENT:/ d # special target
/^\.EXPORT_ALL_VARIABLES:/ d # special target
/^\.NOTPARALLEL:/ d # special target
/^\.ONESHELL:/ d # special target
/^\.POSIX:/ d # special target
/^\.NOEXPORT:/ d # special target
/^\.MAKE:/ d # special target
# The stuff above here describes lines that are not
# explicit targets or not targets other than special ones
# The stuff below here decides whether an explicit target
# should be output.
/^[^#\t:=%]+:([^=]|$)/ { # found target block
h # hold target
d # delete line
}
/^# File is an intermediate prerequisite/ { # nope
s/^.*$//;x # unhold target
d # delete line
}
/^([^#]|$)/ { # end of target block
s/^.*$//;x # unhold target
s/:.*$//p # write current target
d # hide any bugs
}
'
make -npq .DEFAULT 2>/dev/null | sed -n -r "$SCRIPT" \
| sort | uniq
This is a much more complete script than the debian bash completion script because it provides results for generated rules which is what the question asks for and the top voted answer (debian bash completion script at the debian git server) doesn't go far enough.
This is not the original script that I linked to but it is much simpler and is a touch faster.
This is the code for alias based on Todd Hodes solution
alias mtargets='make -qp | awk -F":" "/^[a-zA-Z0-9][^$#\/\t=]*:([^=]|$)/ {split(\$1,A,/ /);for(i in A)print A[i]}"'
This will give a nice output:
make -pn | perl -F: -ane 'print "$F[0]\n" if /^\w+\s*:/' | sort
There is a very nice solution with some sed and egrep magic here: https://gist.github.com/pvdb/777954
Guess I'm a little late to this party, but if you're looking for a specific command you could try
make -qp | grep -v '^# ' | grep -v '^[[:space:]]' | grep --only-matching '^.*:'
This mostly works, although it might still include some non-target stuff like a vpath
directive. If you don't depend on make
's built-in rules, you can use make -qpR
as the first command in the pipeline.
Ruby solution:
`make -qp`.split("\n").select{|l| l =~ /^\w/ && l !~ /=/}.map{|l| l.sub(/:.*/,'')}
Im working on solaris 10 anda turbo C shell. The given solution doesn´t work for my makefile project. even after altering the command line above to tcsh syntax. However, I found out you can get an easy solution using
remake --tasks | grep -v "clean some static output or just grep tabs on start of line" | awk ´{print $1}´
remake version:
remake --version
is
GNU Make 3.82+dbg0.8
Built for sparc-sun-sparc2-9
and some other unimportant version data
I went looking for the same question and came up with this spin:
make -pn | sed -rn '/^[^# \t\.%].*:/p'
This removes all comment lines, pattern rules (lines beginning with tabs), all the intrinsics (example .c.o
and %.o: %.c
patterns).
Found this solution in another thread:
sh -c "cat Makefile | egrep \"^[[:alnum:][:punct:]]{0,}:[[:space:]]{0,}[[:alnum:][:punct:][:space:]]{0,}$\""
You can also add it to your Makefile
:
list:
sh -c "cat Makefile | egrep \"^[[:alnum:][:punct:]]{0,}:[[:space:]]{0,}[[:alnum:][:punct:][:space:]]{0,}$\\""
And execute make list
.
Sure, but when do you want it to spit them out?
To report the name of the target when it runs the rule, put a line in the rule:
foo$(VAR): $(PREREQS)
@echo now making the foo target: $@
do_other_stuff...
To spit them all out at once, you could make a separate PHONY target:
.PHONY: show_vars
show_vars:
@echo foo$(VAR)
@echo bar$(PARAM) blah$(FLAG)
# and so on
And this can be made a prerequisite of your default target:
all: show_vars
...
EDIT:
You want a way to show all possible targets of an arbitrary makefile, which I suppose means non-intrusively. Well...
To do it exactly, and be able to cope with sophisticated makefiles, e.g. involving rules constructed by eval
statements, you'd have to write something close to a Make emulator. Impractical.
To see the targets of the simple rules, you could write a makefile that would act as a makefile scanner, operating on an arbitrary makefile:
- Get all the target names from the makefile using sed.
- `include` the makefile in order to use it to expand variables.
- Use `show_%: ; echo $$*` to print all the targets
This would be an impressive piece of work. Are you sure the goal is worth the effort?
grep ^[a-z].*\:$ Makefile | sed s,:,,g
精彩评论