I am having some trouble with figuring out the default choices of the
values of some of the paths used by the ECL compiler.
I have a fairly vanilla install of ecl, where I think that at least the
relative positions of the installed files concerning ecl are typical. I
find:
> (translate-logical-pathname "SYS:")
#P"/mnt/usb1/scratch/nbruin/4.6.2/local/lib/ecl/"
> (require 'cmp)
;;; Loading #P"/mnt/usb1/scratch/nbruin/4.6.2/local/lib/ecl/cmp.fas"
("CMP")
> c::*ecl-library-directory*
"/mnt/usb1/scratch/nbruin/4.6.2/local/lib/"
> (c::ecl-library-directory)
"/mnt/usb1/scratch/nbruin/4.6.2/local/lib/"
This indicates that the value of c::*ecl-library-directory* was vetted and
found correct. This was done by the test:
(probe-file (merge-pathnames (compile-file-pathname "ecl" :type
#+dlopen :shared-library
#-dlopen :static-library)
c::*ecl-library-directory*))
which succeeds with
#P"/mnt/usb1/scratch/nbruin/4.6.2/local/lib/libecl.so.11.1.1"
However, if I do
> (setf c::*ecl-library-directory* nil)
> (c::ecl-library-directory)
"/mnt/usb1/scratch/nbruin/4.6.2/local/lib/ecl/"
> c::*ecl-library-directory*
"/mnt/usb1/scratch/nbruin/4.6.2/local/lib/ecl/"
I see the system has cooked up another default, namely the directory for
which (probe-file "SYS:BUILD-STAMP") succeeds. However, the file libecl.so
does not reside there!
This means that c::ecl-library-directory, when presented with a value of
c::*ecl-library-directory* it thinks is invalid, overwrites it with a
value that is also invalid by its earlier criterion! That's probably not
as intended.
Shouldn't the value be initialized to the following ?
(namestring (make-pathname :directory (butlast (pathname-directory
(translate-logical-pathname "SYS:")))))
A similar story applies to cc:ecl-include-directory:
> (c::ecl-include-directory)
"/mnt/usb1/scratch/nbruin/4.6.2/local/include/"
which was approved because the following succeeds:
> (probe-file (merge-pathnames "ecl/config.h" c::*ecl-include-directory*))
#P"/mnt/usb1/scratch/nbruin/4.6.2/local/include/ecl/config.h"
Should that fail, then the routine tries if "SYS:" would be a reasonable
candidate via
> (probe-file "SYS:ecl;config.h")
NIL
I don't think that will succeed in any sane installation. The value that
would probably be OK is "SYS:..;..;include", given that:
> (probe-file "SYS:..;..;include;ecl;config.h")
#P"/mnt/usb1/scratch/nbruin/4.6.2/local/lib/ecl-11.1.1/../../include/ecl/config.h"
so probably the thing to try would be:
(namestring (make-pathname :directory (append (butlast (pathname-directory
(translate-logical-pathname "SYS:")) 2) '("include"))))
Most of this doesn't normally matter because ECL captures reasonable
default values for *ecl-library-directory* and *ecl-include-directory* at
build-time.
However, would you be interested in using reasonable defaults derived from
runtime information (being "SYS:") instead? It would mean that ECL
continues to work if its installation is moved after build, as long as the
relative position of headers and libraries is as usual.
Best wishes,
Nils
------------------------------------------------------------------------------
What Every C/C++ and Fortran developer Should Know!
Read this article and learn how Intel has extended the reach of its
next-generation tools to help Windows* and Linux* C/C++ and Fortran
developers boost performance applications - including clusters.
http://p.sf.net/sfu/intel-dev2devmay
_______________________________________________
Ecls-list mailing list
Ecls-list@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/ecls-list
