make errors (missing demolocation)

[cnielsen]

I get the following errors compiling trunk (r37704) though it worked fine a couple days ago.

Errors in:
/src/trunk/db/db.columns
/src/trunk/db/db.connect
/src/trunk/db/db.copy
/src/trunk/db/db.createdb
/src/trunk/db/db.databases
/src/trunk/db/db.describe
/src/trunk/db/db.drivers
/src/trunk/db/db.dropdb
/src/trunk/db/db.droptable
/src/trunk/db/db.execute
/src/trunk/db/db.login
/src/trunk/db/db.select
/src/trunk/db/db.tables
/src/trunk/display/d.barscale
/src/trunk/display/d.colorlist
/src/trunk/display/d.colortable

... there's a lot more but they all have the same error message:

if [ "/src/trunk/dist.i686-pc-mingw32/bin/db.columns.exe" != "" ] ; then GISRC=/src/trunk/dist.i686-pc-mingw32/demolocation/.grassrc70 GISBASE=/src/trunk/dist.i686-pc-mingw32 PATH="/src/trunk/dist.i686-pc-mingw32/bin:$PATH" PYTHONPATH="c:\programs\gis\osgeo4w\usr\src\trunk\dist.i686-pc-mingw32\etc\python;$PYTHONPATH" PATH="/src/trunk/dist.i686-pc-mingw32/bin:/src/trunk/dist.i686-pc-mingw32/lib:/src/trunk/dist.i686-pc-mingw32/lib:.:/usr/local/bin:/mingw/bin:/bin:/osgeo4w/bin:/c/Windows/system32:/c/Windows:/c/Windows/System32/Wbem:/c/PROGRA~1/COMMON~1/ULEADS~1/MPEG:/c/Programs/GIS/Subversion:/c/Program Files/QuickTime/QTSystem/:/osgeo4w/apps/grass/bin:/bin:/c/Programs/Subversion" LC_ALL=C /src/trunk/dist.i686-pc-mingw32/bin/db.columns.exe --html-description < /dev/null | grep -v '</body>\|</html>' > db.columns.tmp.html ; fi
access: No such file or directory
ERROR: LOCATION << /src/trunk/dist.i686-pc-mingw32/demolocation >> not
       available
make: *** [db.columns.tmp.html] Error 1
rm db.columns.tmp.html

[glynn]

Replying to cnielsen:

I get the following errors compiling trunk (r37704) though it worked fine a couple days ago.

I've been making quite a lot of changes to the build system over the last couple of days.

... there's a lot more but they all have the same error message:

ERROR: LOCATION << /src/trunk/dist.i686-pc-mingw32/demolocation >> not
       available

Try this:

Index: include/Make/Grass.make
===================================================================
--- include/Make/Grass.make	(revision 37704)
+++ include/Make/Grass.make	(working copy)
@@ -48,7 +48,6 @@
 LIB_RUNTIME_DIR = $(GISBASE)/lib
 
 RUN_GISRC       = $(GISBASE)/demolocation/.grassrc$(GRASS_VERSION_MAJOR)$(GRASS_VERSION_MINOR)
-RUN_GISBASE     = $(GISBASE)
 
 #########################################################################
 # these define the various directories which contain GRASS programs
Index: include/Make/Platform.make.in
===================================================================
--- include/Make/Platform.make.in	(revision 37704)
+++ include/Make/Platform.make.in	(working copy)
@@ -41,6 +41,7 @@
 INST_DIR            = @INSTDIR@
 
 GRASS_HOME          = @GRASS_HOME@
+RUN_GISBASE         = @GISBASE@
 
 GRASS_VERSION_MAJOR = @GRASS_VERSION_MAJOR@
 GRASS_VERSION_MINOR = @GRASS_VERSION_MINOR@

[cnielsen]

New error:

Tue Jun 2 14:50:20 EDT 2009: STARTING make
include/Make/Vars.make:1: include/Make/Platform.make: No such file or directory
make: *** No rule to make target `include/Make/Platform.make'.  Stop.
include/Make/Vars.make:1: include/Make/Platform.make: No such file or directory
make: *** No rule to make target `include/Make/Platform.make'.  Stop.

[glynn]

Replying to cnielsen:

New error:

include/Make/Vars.make:1: include/Make/Platform.make: No such file or directory

Can you provide more detail on where this is originating from, i.e. which Makefile is generating this error?

The error implies that MODULE_TOPDIR isn't being set.

[cnielsen]

I can't seem to reproduce the second error, meaning probably my error. demolocation error fixed with the patch.

[cnielsen]

Sorry Glynn, I had rolled back to an earlier version and that is why the demolocation error seemed fixed. In fact the patch had no effect, the error is still present.

/src/trunk/db/db.columns
$ make
if [ "/src/trunk/dist.i686-pc-mingw32/bin/db.columns.exe" != "" ] ; then GISRC=/src/trunk/dist.i686-pc-mingw32/demolocation/.grassrc70 GISBASE=c:/programs/gis/osgeo4w/usr/src/trunk/dist.i686-pc-mingw32 PATH="/src/trunk/dist.i686-pc-mingw32/bin:$PATH" PYTHONPATH="c:\programs\gis\osgeo4w\usr\src\trunk\dist.i686-pc-mingw32\etc\python;$PYTHONPATH" PATH="/src/trunk/dist.i686-pc-mingw32/bin:/src/trunk/dist.i686-pc-mingw32/lib:/src/trunk/dist.i686-pc-mingw32/lib:.:/usr/local/bin:/mingw/bin:/bin:/osgeo4w/bin:/c/Windows/system32:/c/Windows:/c/Windows/System32/Wbem:/c/PROGRA~1/COMMON~1/ULEADS~1/MPEG:/c/Programs/GIS/Subversion:/c/Programs/Online/QuickTime/QTSystem/:/osgeo4w/apps/grass/bin:/bin:/c/Programs/Subversion" LC_ALL=C /src/trunk/dist.i686-pc-mingw32/bin/db.columns.exe --html-description < /dev/null | grep -v '</body>\|</html>' > db.columns.tmp.html ; fi
access: No such file or directory
ERROR: LOCATION << /src/trunk/dist.i686-pc-mingw32/demolocation >> not
       available
make: *** [db.columns.tmp.html] Error 1
rm db.columns.tmp.html

[glynn]

Replying to cnielsen:

Sorry Glynn, I had rolled back to an earlier version and that is why the demolocation error seemed fixed. In fact the patch had no effect, the error is still present.

ERROR: LOCATION << /src/trunk/dist.i686-pc-mingw32/demolocation >> not
       available

This should be fixed in r37808.

[hamish]

Hi,

I can reproduce something similar on amd64 Linux with r37899. (pretty new Debian/Lenny)

the problem seems to be that the system type is not detected properly ("unknown" or ""), and build dirs end up called bin./ and dist./.

there is: "config.status.x86_64-unknown-linux-gnu"

for releasebranch_6_4 on the same machine I get: bin.x86_64-unknown-linux-gnu/ and dist.x86_64-unknown-linux-gnu/ and it builds ok.

error is:

[...]grass/svn/trunk/dist./bin/db.login --html-description < /dev/null | grep -v '</body>\|</html>' > db.login.tmp.html ; fi
access: No such file or directory
ERROR: LOCATION <<
       /usr/src/grass/svn/trunk/dist.unknown/demolocation >> not
       available
make[2]: *** [db.login.tmp.html] Error 1
rm db.login.tmp.html
make[2]: Leaving directory `/usr/src/grass/svn/trunk/db/db.login'
make[1]: Leaving directory `/usr/src/grass/svn/trunk/db'

in the above note dist. != dist.unknown.

I had a look but couldn't see where in configure's uname magic it determines the "unknown" type. Seems to be related to cross-compiling.

?, Hamish

[hamish]

same thing happens with develbranch6. relbr64 builds ok.

Platform.make's ARCH = line is empty.

Hamish

[glynn]

Replying to hamish:

I can reproduce something similar on amd64 Linux with r37899. (pretty new Debian/Lenny)

the problem seems to be that the system type is not detected properly ("unknown" or ""), and build dirs end up called bin./ and dist./.

System detection is up to config.guess; we might need a newer version.

there is: "config.status.x86_64-unknown-linux-gnu"

Are you sure that this isn't a leftover from a previous build?

for releasebranch_6_4 on the same machine I get: bin.x86_64-unknown-linux-gnu/ and dist.x86_64-unknown-linux-gnu/ and it builds ok.

error is:

ERROR: LOCATION << /usr/src/grass/svn/trunk/dist.unknown/demolocation >> not available

in the above note dist. != dist.unknown.

I had a look but couldn't see where in configure's uname magic it determines the "unknown" type. Seems to be related to cross-compiling.

# Set ARCH

ARCH=
if test -z "${host}"; then
ARCH="unknown"
else
ARCH="${host}"
fi

Note that Platform.make.in has:

ARCH                = @host@

Although configure.in has:

AC_SUBST(ARCH)

ARCH isn't used in Platform.make.

Also, SC_CONFIG_FLAGS uses uname (which means that cross-compiling can't work), while everything else uses $host.

Which reminds me that SC_CONFIG_FLAGS should have been discarded for 7.0.

[hamish]

Replying to glynn:

System detection is up to config.guess; we might need a newer version.

there is: "config.status.x86_64-unknown-linux-gnu"

Are you sure that this isn't a leftover from a previous build?

that's in the relbr6 (which builds). In devbr6 and trunk on the same system it is called config.status.unknown.

> # Set ARCH
> 
> ARCH=
> if test -z "${host}"; then
> ARCH="unknown"
> else
> ARCH="${host}"
> fi

Note that Platform.make.in has:

> ARCH                = @host@

Although configure.in has:

> AC_SUBST(ARCH)

ARCH isn't used in Platform.make.

Also, SC_CONFIG_FLAGS uses uname (which means that cross-compiling can't work), while everything else uses $host.

Which reminds me that SC_CONFIG_FLAGS should have been discarded for 7.0.

ok, following config.guess:

timestamp='2008-04-14'
...
UNAME_MACHINE=`(uname -m) 2>/dev/null` || UNAME_MACHINE=unknown
UNAME_RELEASE=`(uname -r) 2>/dev/null` || UNAME_RELEASE=unknown
UNAME_SYSTEM=`(uname -s) 2>/dev/null`  || UNAME_SYSTEM=unknown
UNAME_VERSION=`(uname -v) 2>/dev/null` || UNAME_VERSION=unknown

echo "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}"  
x86_64:Linux:2.6.26-2-amd64:#1 SMP Thu May 28 21:28:49 UTC 2009

...leads to
    x86_64:Linux:*:*)
        echo x86_64-unknown-linux-gnu
        exit ;;

I don't see where $host is set.

trunk$ ./configure reports:

checking host system type... x86_64-unknown-linux-gnu
...
Copying config.status to config.status.x86_64-unknown-linux-gnu

GRASS is now configured for:  x86_64-unknown-linux-gnu

Platform.make now has:

ARCH                = x86_64-unknown-linux-gnu

weird, this time it got it. more tests tomorrow.

Hamish

[glynn]

Replying to hamish:

ok, following config.guess:

I don't see where $host is set.

It's set by AC_CANONICAL_HOST from the output of config.guess.

[hamish]

I think I figured out how I was getting "config.status.unknown".

it was because I messed up the ./configure line:

CFLAGS="-ggdb -march=amdfam10 -Wall -Werror-implicit-function-declaration" \
  ./configure \
    --with-tcltk-includes=/usr/include/tcl8.5 --with-cairo \
    --with-motif --with-python=/usr/bin/python2.5-config \
    --with-readline --with-cxx --with-odbc --with-sqlite \
    --with-freetype --with-freetype-includes=/usr/include/freetype2 \
    --without-postgres --with-proj-share=/usr/share/proj \
    --enable-64bit --with-pthread \
    --with-wxwidgets=/usr/lib/wx/config/gtk2-unicode-release-2.8 \
    \ #--with-ffmpeg --with-ffmpeg-includes=/usr/include/ffmpeg \
    --with-geos \
    2>&1 | tee config_log.txt

\ #comment is a problem.. causes this to happen when you run ./configure:

configure: warning:  #--with-ffmpeg: invalid host type
checking host system type... Invalid configuration `#--with-ffmpeg': machine `#--with' not recognized

checking for gcc... gcc

I don't really understand why ./configure continues weirdly, but if I get rid of the \ #--with-ffmpeg line it all builds ok.

sorry for hijacking this bug with a stupid mistake, Hamish

[cnielsen]

Replying to glynn:

This should be fixed in r37808.

Error resolved, closing ticket. (Good to see Hamish's error is resolved too)

[glynn]

Replying to hamish:

\ #comment is a problem.. causes this to happen when you run ./configure:

configure: warning:  #--with-ffmpeg: invalid host type

I don't really understand why ./configure continues weirdly, but if I get rid of the \ #--with-ffmpeg line it all builds ok.

bash treats "\ "" as a literal space, not a word separator. This means that the following # isn't at the beginning of a word, so it's treated as a literal # character rather than starting a comment. The end result is an argument of " #--with-ffmpeg". As it's the first argument which doesn't begin with a dash, it's treated as the host type.