Re: building executables with pattern rules

[Gavin Smith]

On Mon, Jul 14, 2014 at 6:49 PM, Will Estes <address@hidden> wrote:
> I have the following in my Makefile.am, and automake is not using the rule to 
> build the executable.
>
> tableopts_nr%.c: tableopts.l4
>         $(FLEX) -P $(subst -,_,$(*F)) $* -o $@ $<
>
> tableopts_nr%.$(OBJEXT): tableopts_nr%.c
>         $(COMPILE) -c -o $@ $<
>
> tableopts_nr%.tableopts$(EXEEXT): tableopts_nr.$(OBJEXT)
>         $(LINK) $^
>
> The % is matching various things like -Ca and so on that are options to flex 
> for different kinds of tables that flex can produce. This is in the test 
> suite and I've generated the long list of _SOURCES variables and the programs 
> are listed in CHECK_PROGRAMS and TESTS.
>
> The .tableopts  suffix is present because I'm using the parallel test harness 
> to distinguish between various kinds of tests in the flex test suite that 
> need different kinds of arguments when they're run.
>
> THe output from trying to build the program with the Makefile.am as above 
> looks like this:
>
> /bin/bash ../libtool  --tag=CC   --mode=link gcc  -g -O2   -o 
> tableopts_nr-Ca.tableopts   -lm
> libtool: link: gcc -g -O2 -o tableopts_nr-Ca.tableopts  -lm
> /usr/lib/gcc/x86_64-linux-gnu/4.7/../../../x86_64-linux-gnu/crt1.o: In 
> function `_start':
> (.text+0x20): undefined reference to `main'
> collect2: error: ld returned 1 exit status
>
> As a test, I did try explicitly listing tableopts_nr-Ca.tableopts$(EXEEXT) as 
> a target and make did build the program correctly.
>
> How do I get automake to recognize my pattern rule? Or, if there's another 
> way to do this that's easier / better, what is that?

Try checking your generated "Makefile.in" and "Makefile" files to see
if automake has copied these rules through. If not, it could be
because automake doesn't recognize GNU make extensions (pattern rules
with % don't exist in some other make programs). However, if it did
succeed in passing them through, it could be a problem with your
rules. I thought it was strange that in the rule

> tableopts_nr%.tableopts$(EXEEXT): tableopts_nr.$(OBJEXT)
>         $(LINK) $^

the object file doesn't have a % in it (so the same file will be used
for all matches of this pattern), but in


> tableopts_nr%.$(OBJEXT): tableopts_nr%.c
>         $(COMPILE) -c -o $@ $<

there is a % there. Try making a simple Makefile to check the logic of
your rules first and then try to incorporate it in your Makefile.am.