Re: problem with hardcode_libdir_flag_spec

[Werner LEMBERG]

This is a shy `ping' :-)


    Werner


----------------------------------------------------------------------


> [libtool 2.4.2]
> 
> Folks,
> 
> 
> let's assume that someone is doing
> 
>   AC_SUBST([libdir])
>   AC_SUBST([wl])
>   AC_SUBST([hardcode_libdir_flag_spec])
> 
> and later tries to say
> 
>   libdir = @libdir@
>   wl = @wl@
>   hardcode_libdir_flag_spec = @hardcode_libdir_flag_spec@
> 
> within a Makefile, as can be done with other such variables.  [The
> idea is to construct a proper `-R' linker argument for a `--libs'
> option of a `foo-config' file generated at compile time, but this is a
> different issue.]
> 
> However, this causes a very unpleasant surprise since the above
> snippet gets converted to
> 
>   libdir = /foo/bar
>   wl = -Wl,
>   hardcode_libdir_flag_spec = '${wl}-rpath ${wl}$libdir'
> 
> Note the `$libdir' string which gets interpreted as variable `$l'
> followed by `ibdir' within a Makefile...
> 
> Now my questions.
> 
>   . Is it a bug?  It should be trivial IMHO to use ${libdir} instead
>     to avoid incorrect Makefile variable expansion.  I can imagine
>     that other variables (even in autoconf or automake) are similarly
>     affected.
> 
>   . Is it correct usage?  I mean, is there a list of autotools
>     variables which must not be used within a Makefile (after
>     AC_SUBST)?
> 
>   . How can I circumvent the problem with the current libtool version?
> 
> 
>   Werner