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

[Maxim Nikulin]

Dear John,

John Graham-Cumming wrote:
> Specifically this happens in your case 1 (in the current directory) 
> which you can get around by using $(CURDIR):
>
>     bar: -lfoo ; touch $@
>     $(CURDIR)/libfoo.so: ; touch $@
>     vpath lib%.so $(CURDIR)

Thank you very much. It's a bright idea. It seems it can help with 
another my problem too:

Maxim Nikulin wrote:
> 3. Let's consider deep complicated recursive build (I wouldn't argue
> about non-recursive approach and its advantages). Every package install
> headers and libraries into intermediate directories so they are
> available for other packages. I tried to speed up rebuilding by
> minimizing of copy operations. The aim is not to copy a file in the
> intermediate directory if it's not outdated. The rule
>
> /intermediate/dir/libfoo.so: libfoo.so
>     $(INSTALL) $< $@
>
> leads to complains on circular dependence if /intermediate/dir is in vpath. 
(To clarify I should add 'if libfoo.so isn't built yet')

If the absolute path is specified before libfoo.so the problem 
disappears. However real life is more complicated. I have to invoke 
alien makefiles during the build. They don't know about intermediate 
installation so I use 'cp --preserve=timestamps' as INSTALL. In case of 
such INSTALL in my rules make sometimes decides that it should reinstall 
file although it hasn't been updated. I think it's due Linux Virtual 
File System has subsecond precision of timestamps, but ext3 file system 
stores second only.

--
Maxim Nikulin