Re: VPATH/vpath: does it work for -l<libname> prereqs?

[Matt England]

At 11/22/2005 01:09 PM, John Graham-Cumming wrote:
> Yes, but there's special handling of this syntax detailed in the GNU Make 
> manual section "Directory Search for Link Libraries" and I quote:
>   "When a prerequisite's name has the form `-lNAME', [...]

Yes, I'm familiar with these rules, I've been studying them intently at:
http://www.gnu.org/software/make/manual/html_chapter/make_4.html#SEC40
Unfortunately, the system is not behaving as per my interpretation of the 
above.  My interpretation may be faulty, but I'm not sure.

> You might like to post a snippet of Makefile so that people here can help 
> you debug.

Yep, I'm working to make a suitably-isolated portion of or entire Makefile 
so I can pass it to the list, it just takes some work to do this to make it 
understandable by the group.  Stay tuned...if I can't figure it out.

Thanks for passing along the examples.  These examples all currently match 
my understanding.

> Now one thing to watch for is if your library is a .a and has the same 
> name as a .so the search order used by GNU Make (see above) will cause the 
> .so to be found and used first.  If that's a problem then you can either 
> stop using the -l syntax, or you can fiddle with .LIBPATTERNS so that the 
> search happens in the opposite order.

I do have .a and .so files of same names in my paths (my build procedures 
produce both), and I'm already setting .LIBPATTERNS as per desired behavior.

...and all this stuff works...IF I keep the lib dur in my current directory 
tree and/or use only ONE dir path in vpath/VPATH (I need multiple lib dir 
paths because I have multiple lib directories).

Any known problems vpath/VPATH and multiple lib dirs?

-Matt