Re: RFE: support per-program noinst_HEADERS

[Chris Pickett]

Hi Ralf,

Ralf Wildenhues wrote:
> Hi Chris,
>
> * Chris Pickett wrote on Wed, Mar 12, 2008 at 07:50:46PM CET:
> > I would like to be able to use noinst_prog_HEADERS, but with 1.10.1 I  
> > get a complaint about noinst_progdir not being defined:
>
> You can just put all your not-to-be installed headers into the *_SOURCES
> alongside the .c/.cc/... files.  *_HEADERS really is necessary only for
> headers that are to be installed.

That doesn't work for me.  I am #include'ing .c files, and I do not want 
objects built for them.  This works fine with noinst_HEADERS for a 
single library or program:

if WPA
  libspmt_la_SOURCES = private.c
  private.c: $(noinst_HEADERS)
  noinst_HEADERS = $(ACTUAL_SOURCES)
else # !WPA 



  libspmt_la_SOURCES = $(ACTUAL_SOURCES)
endif # !WPA 




where private.c just #include's all of the other .c files in the program.

However, I recently got frustrated trying to run tests against the 
library.  I am creating one binary for each module in the library and 
running them all with make check.  Before, I was including the library 
with LDADD, and so if a single file in the library changed while trying 
to make some test pass then all of the tests would get recompiled.  So, 
now I generate a list of src/ .c files that each tests/ .c files needs 
to include in order to create a proper binary, and #include them all. 
Disgusting perhaps, but my turnaround time after making a change in the 
library to try and make a test pass is about 10 seconds now.

I did play with other methods, like adding the objects to prog_LDADD, 
but I couldn't get them to work.  It might be because I'm in tests/ and 
the objects I want are in src/.  Also, I was having problems just 
listing things in module_SOURCES because the files are identically 
named: I have src/module.c and tests/module.c

Cheers,
Chris

P.S.  I also did get this to work using prog_DEPENDENCIES, but the 
manual warns that it is not for source code.