RE: [Gcl-devel] Re: delayed pathname.d patch

[Mike Thomas]

Hi Camm.

Being brief here as in a hurry again - should know better but too bad!

| OK, you've apparently discovered that it is the missing winnt in
| *features* that was causing your problem.
|
| I have a different suggestion that your recent commit to main.c:  What
| is your canonical host string output by configure?  I'm hoping one of
| either the third or fourth components is either winnt or win32.

Unfortunately not, hence the nature of the patch - in fact host on Mingw32
is reported as the compiler name rather than the OS - a bit of a bummer but
a practical result of the hosting of Unix tools under Cygwin. MSYS and
others etc.


| All
| that is used to my understanding is winnt, and this only in
| gcl_cmpmain.lsp, so lets try to stick to one item, preferably
| autodetected by configure like on other systems, and make sure
| gcl_cmpmain.lsp matches.
|
| Separately, there appears to be a straightforward fix to what appears
| to have been the original motivation behind one section of the #+winnt
| code in gcl_cmpmain.lsp.  I've just committed some changes which
| should allow compiles to work without the following block:  could you
| please confirm?

Sorry but doesn't work - don't mod as I will follow up.  Had to laugh as I
made almost space-for-space excatkly same patches on other computer sown to
variable names.  These changes are appropriate however and should remain.
Problems occur for other reasonas and I will correct those soon.

|
|    #+(or dos winnt)
|       (format nil "~a -I~a ~a ~a -c -w ~a -o ~a"
|             *cc*
|             (concatenate 'string si::*system-directory* "../h")
|             (if (and (boundp '*c-debug*) *c-debug*) " -g " "")
|             (case *speed*
|                   (3 *opt-three* )
|                   (2 *opt-two*)
|                   (t ""))
|             (namestring (make-pathname  :type "c" :defaults (first args)))
|             (namestring (make-pathname  :type "o" :defaults (first args)))
|             )


Unfortunately there are subtleties (eg single quotes in system calls and ~
in shortened file names) which are OS specific versus options/API's which
are compiler specific so I don't believe this is a good idea - I have
purposely tried to allow for the possibility of someone in future using in
particular either Cygwin or Visual C++.  Also must allow for WinCE and
_WIN64.

| I also don't understand the rationale behind these patches to
| configure.in:
|
| case "$use" in
|   *mingw*)
|       TFPFLAG="-fomit-frame-pointer";;
|   *)
|       TFPFLAG="-fomit-frame-pointer";;
| esac
|
| ...
|
|       case "$use" in
|         *mingw*)
|               TO3FLAGS="-O3 $TFPFLAG";;
|         *)
|               TO3FLAGS="-O3 $TFPFLAG";;
|       esac
|
|
| Can't we revert these?

Prefer not for testing change convenience unless you strongly object.

Cheers

Mike Thomas.


|
| Take care,
|
|
| --
| Camm Maguire                                          address@hidden
| ==========================================================================
| "The earth is but one country, and mankind its citizens."  --  Baha'u'llah
|
|
| _______________________________________________
| Gcl-devel mailing list
| address@hidden
| http://mail.gnu.org/mailman/listinfo/gcl-devel
|
|