RE: [Gcl-devel] 2.6.2

[Mike Thomas]

Hi Camm.

| > For some reason gcl 2.6.2 (and current 2.6.1) has gone backwards re gcc
| > 3.4.0 on Windows, not even being able to compile.  This is
| apart from the
| > need to patch o/makefile to workaround the -I hanging directory name
| > separator bug as follows:
| >
| > ====================================================================
| > $ cvs diff o/makefile
|
| Mike, I'm confused as to why this is needed -- we've always defined
| directory variables with a trailing slash.

In o/makefile we are actually harmlessly inconsistent eg:

HDIR    = ../h/
CDIR    = ./
MPDIR   = ../mp

#OFLAG  = -O
DEFS    = -I../h -I../gcl-tk
ODIR    = .

and even our handling of HDIR is inconsistent in the same makefile macro:

CMPINCLUDE_FILES=${HDIR}compbas.h ${HDIR}enum.h ${HDIR}object.h ${HDIR}vs.h
\
        ${HDIR}bds.h ${HDIR}frame.h \
        ${HDIR}lex.h ${HDIR}eval.h    ${HDIR}funlink.h \
        ${HDIR}att_ext.h ${HDIR}compbas2.h ${HDIR}/compat.h ${HDIR}/cmponly.h



|  What exactly is the ' -I
| hanging directory name separator' bug and its symptoms?

I reported a few weeks ago problems with the new_decl.h:

============================================================================
=============
make[1]: Entering directory `/c/cvs/stable/gcl-2.6.1/o'
gcc -c -g -Wall -DVOL=volatile -fsigned-char -fwritable-strings -pipe -fno-z
ero-initialized-in-bss -mms-bitfields -g  -mcpu=i386 -march=i386  -I/c/cvs/s
table/gcl-2.6.1/o -I../h -I../gcl-tk main.c
cc1.exe: note: -fwritable-strings is deprecated; see documentation for
details In file included from ../h/include.h:74,                  from
main.c:49: ../h/../h/new_decl.h: In function `fLbye':
../h/../h/new_decl.h:3: error: syntax error before '{' token

============================================================================
=============


MinGW32 gcc 3.4.0 on Windows has a bug in which a trailing / on a -I
directory stopped the compiler from finding files in that directory.


|
| As for the 3.4.0, the way forward is in two directions:
|
| 1) Add -g to the CFLAGS, and give me a gdb backtrace when loading
|    gcl_setf.lisp.
|
| 2) Take dated CVS snapshots going back sometime in the last month to
|    narrow down the change in question.  I'm asusming the gcc/lib
|    install hasn't changed in the intervening time.

I'm going to leave this for now - there are too many other problems and I
think I'll just specify gcc 3.3.1 on Windows for GCL 2.6.2.  Note that with
all the Unix environment and build tool emulation machinery needed to build
GCL on Windows a complete build takes about half an hour - this kind of
testing is awfully time consuming.

| and a question:
|
| 3) Does the 3.4.0 compile go through with --enable-debug?

I believe it does but my memory of a couple of days ago is now fast fading.

Have a good weekend

Mike Thomas.