"include/linux" in mapscript/java Makefile.in

[havard.tveite@…]

Problem compiling java mapscript from 20040203 CVS.
The mapscript/java/Makefile.in contains the line:
JAVA_INCLUDE=-I${JAVA_HOME}/include -I${JAVA_HOME}/include/linux

My OpSys is SunOS 5.7 (solaris). My SWIG installation contains
an include/solaris directory (and not include/linux).

[havard.tveite@…]

Sorry, but what I meant to say was that my *Java* (not SWIG)
installation contains an include/solaris directory.

[sgillies@…]

Not having solaris at my disposal, I won't be able to conclusively test ... but
I will try to implement some platform detection.  Should be able to do it in the
same way as done in the main mapserver/Makefile.in.

[sgillies@…]

Setting target to 4.6

[sgillies@…]

CC'ing Umberto. Any ideas, Umberto?

[unicoletti]

I hacked configure.in and added a --with-os-name parameter (give the 'solaris'
value) to configure so that a different include patch can be used. This is far
from ideal (We should be using the AC_CANONICAL_SYSTEM macro instead, but that
requires install-sh so I went for a home-made solution).

Please try the attached patch and report if it works.
The output of
uname -a

and 

uname -s

on solaris would be handy too.

[unicoletti]

Add os-name parameter to search a different path for includes

[havard.tveite@…]

I have tried the patch.
My first problem was that my "patch" did not like the file (too long lines?):

patch output:
*********************************************************************
Hmm...  Looks like a unified diff to me...
The text leading up to this was:
--------------------------
|Index: configure
|===================================================================
|RCS file: /data2/cvsroot/mapserver/configure,v
|retrieving revision 1.135
|diff -u -r1.135 configure
|--- configure  27 Feb 2006 20:22:06 -0000      1.135
|+++ configure  28 Feb 2006 16:14:59 -0000
--------------------------
Patching file configure using Plan A...
patch: **** malformed patch at line 13: Y_CONFIG MYGIS_ENABLED MYGIS_INC
MYGIS_LIB ORACLESPATIAL_ENABLED ORACLESPATIAL_INC ORACLESPATIAL_LIB MING_ENABLED
MING_INC MING_LIB WMS_SVR_ENABLED WFS_SVR_ENABLED WCS_SVR_ENABLED LIBCURL_CONFIG
WMS_LYR_ENABLED WFS_LYR_ENABLED CURL_INC CURL_LIB FASTCGI_ENABLED FASTCGI_INC
FASTCGI_LIB HTTPD MS_INC MS_LIB IGNORE_MISSING_DATA USE_POINT_Z_M DEBUG_FLAGS
LD_SHARED SO_EXT PHP_VERSION_FLAG PHP_SRC_DIR PHP_CC PHP_LD PHP_REGEX_INC
PHP_REGEX_OBJ USE_PHP_REGEX MAKE_PHPMS MAKE_PHPMS_CLEAN APACHE_INC ALL_ENABLED
ALL_INC ALL_LIB ALL_STATIC_LIB RPATHS LIBOBJS LTLIBOBJS'
*********************************************************************

I did the change to ac_subst_vars (inserted "OS_INCLUDE_DIR") manually and
removed the diff from the patch. The rest of the patch was OK.


Problem encountered: 

The section:
****************************************************
OUR_OSNAME=`uname -s`
case "${OUR_OSNAME}" in
Llinux) OS_INCLUDE_DIR="linux" ;;
sSolaris*) OS_INCLUDE_DIR="solaris" ;;
Wwin*) OS_INCLUDE_DIR="win32" ;;
*) OS_INCLUDE_DIR="linux" ;;
****************************************************
fails to discover my Solaris 2.7, since uname -s reports
"SunOS"
You also asked for the output of uname -a:
Solaris 2.7: "SunOS hostname 5.7 Generic_106541-34 sun4u sparc SUNW,Ultra-4"
Solaris 2.5: "SunOS hostname 5.5.1 Generic_103640-40 sun4u sparc SUNW,Ultra-1"

by running configure without "--with-os-name", configure
reported:
"checking if --with-os-name specified... no, but it seems to be linux"

using --with-os-name=solaris, configure reports:
"checking if --with-os-name specified... yes (solaris)"
(as expected)

In the last case, the mapscript/java/Makefile was updated
correctly, and compilation went better. But the compilation
failed due to some problems with m and z in mapscript_wrap.c
(i did configure with --enable-point-z-m):
*****************************************************************
mapscript_wrap.c: In function `pointObj_setXY':
mapscript_wrap.c:2498: error: structure has no member named `z'
mapscript_wrap.c:2499: error: structure has no member named `m'
mapscript_wrap.c: In function `pointObj_setXYZ':
mapscript_wrap.c:2506: error: structure has no member named `z'
mapscript_wrap.c:2507: error: structure has no member named `m'
mapscript_wrap.c: In function `pointObj_setXYZM':
mapscript_wrap.c:2513: error: structure has no member named `z'
mapscript_wrap.c:2514: error: structure has no member named `m'
mapscript_wrap.c: In function `pointObj_toString':
mapscript_wrap.c:2521: error: structure has no member named `m'
mapscript_wrap.c: In function `new_pointObj__SWIG_0':
mapscript_wrap.c:2535: error: structure has no member named `z'
mapscript_wrap.c:2536: error: structure has no member named `m'
mapscript_wrap.c: In function `new_pointObj__SWIG_1':
mapscript_wrap.c:2545: error: structure has no member named `z'
mapscript_wrap.c:2546: error: structure has no member named `m'
mapscript_wrap.c: In function
`Java_edu_umn_gis_mapscript_mapscriptJNI_set_1pointObj_1z':
mapscript_wrap.c:22422: error: structure has no member named `z'
mapscript_wrap.c: In function
`Java_edu_umn_gis_mapscript_mapscriptJNI_get_1pointObj_1z':
mapscript_wrap.c:22435: error: structure has no member named `z'
mapscript_wrap.c: In function
`Java_edu_umn_gis_mapscript_mapscriptJNI_set_1pointObj_1m':
mapscript_wrap.c:22450: error: structure has no member named `m'
mapscript_wrap.c: In function
`Java_edu_umn_gis_mapscript_mapscriptJNI_get_1pointObj_1m':
mapscript_wrap.c:22463: error: structure has no member named `m'
make: *** [mapscript_so] Error 1
*****************************************************************

[unicoletti]

In fact the patch has a rather long line. What patch are you using (gnu or sun)?

As for the os detection code shortly after I have attached the patch I found in
a web site a sample uname -s output. I already have updated the patch and will
provide a new one shortly.

With regard to --enable-point-z-m I recall that this has always been an issue
with Java mapscript, try to disable it or if you absolutely need it open a new bug.

[havard.tveite@…]

I used Solaris patch version 2.1

Regarding the z-m issue. What I meant to say was that I did *not* configure with
--enable-point-z-m. Sorry!

[havard.tveite@…]

In the mapscript/java/Makefile, USE_POINT_Z_M is defined
by configure according to the presence of --enable-point-z-m
in the configure parameters, but it not used in the Makefile.

I included "$(USE_POINT_Z_M)" among the other options in
CCFLAGS in the Makefile, and this made the compilation run
fine (in the mean time, I had also configured and compiled
Mapserver with --enable-point-z-m).
Compilation still fails if the --enable-point-z-m is not used
in configure (the use of z and m is not "protected" by ifdefs
in mapscript_wrap.c).

[unicoletti]

OS detection code for configure and Java mapscript

[unicoletti]

Please open a new bug issue for point-z-m and add me to the cc-list

[havard.tveite@…]

The last patch does not update configure with the
"SunOS" stuff, only configure.in.
I ran autoconf on configure.in, and configure is still
not able to detect SunOS. It seems as if the pattern:
"sSunOS*" does not produce a hit.
I changed the pattern to "SunOS*", and that worked.
According to man sh, "[sS]unOS*" should also work.

[unicoletti]

OS detection code for configure and Java mapscript

[havard.tveite@…]

The patch is OK here now.
In order to make mapscript/java, the Makefile.in must be modified
to include $(USE_POINT_Z_M) in CCFLAGS.

I have created a bug (id=1695) for the USE_POINT_Z_M issues with
Java mapscript.

[sgillies@…]

Keep in mind that there are actually very few use cases for Z and M in
MapServer. The MapServer program itself can not use these values. And they also
cause a 10-15% performance loss (check Bugzilla for details). I do not recommend
enabling them by default.

[unicoletti]

The proposed patch has been applied to cvs HEAD