Syntax error in javapre.i

[devidfil@…]

Trying to build some versions of CS in Ubuntu and Debian, CS fails to build with this error:

cc1plus: error: unrecognized command line option "-Wno-long-double"

g++ -c -o ./out/linuxx86/optimize/plugins/video/render3d/shader/shadercompiler/xmlshader/cpi/expparser.o -I. -I./include -I./include -Wall -g -O2 -pipe -Wall -g -O2 -Wall -Wno-unknown-pragmas -fvisibility=hidden -I/usr/local/include -Wno-long-double -g -O2 -g -O2 -fno-exceptions -fvisibility-inlines-hidden -O3 -fomit-frame-pointer -ffast-math -g2 -fPIC  plugins/video/render3d/shader/shadercompiler/xmlshader/cpi/expparser.cpp

...skipped xmlshader.so for lack of <plugins!video!render3d!shader!shadercompiler!xmlshader!>shader.o...

mkdir -p "./out/linuxx86/optimize/plugins/video/render3d/null"


g++ -c -o ./out/linuxx86/optimize/plugins/video/render3d/null/null_render3d.o -I. -I./include -I./include -Wall -g -O2 -pipe -Wall -g -O2 -Wall -Wno-unknown-pragmas -fvisibility=hidden -I/usr/local/include -Wno-long-double -g -O2 -g -O2 -fno-exceptions -fvisibility-inlines-hidden -O3 -fomit-frame-pointer -ffast-math -g2 -fPIC  plugins/video/render3d/null/null_render3d.cpp


g++ -c -o ./out/linuxx86/optimize/plugins/video/render3d/null/null_txt.o -I. -I./include -I./include -Wall -g -O2 -pipe -Wall -g -O2 -Wall -Wno-unknown-pragmas -fvisibility=hidden -I/usr/local/include -Wno-long-double -g -O2 -g -O2 -fno-exceptions -fvisibility-inlines-hidden -O3 -fomit-frame-pointer -ffast-math -g2 -fPIC  plugins/video/render3d/null/null_txt.cpp


g++ -c -o ./out/linuxx86/optimize/plugins/video/render3d/null/null_pix.o -I. -I./include -I./include -Wall -g -O2 -pipe -Wall -g -O2 -Wall -Wno-unknown-pragmas -fvisibility=hidden -I/usr/local/include -Wno-long-double -g -O2 -g -O2 -fno-exceptions -fvisibility-inlines-hidden -O3 -fomit-frame-pointer -ffast-math -g2 -fPIC  plugins/video/render3d/null/null_pix.cpp

...skipped null3d.so for lack of libcrystalspace-1.2.a...

mkdir -p "./out/linuxx86/optimize/plugins/video/cursor"


g++ -c -o ./out/linuxx86/optimize/plugins/video/cursor/cursor.o -I. -I./include -I./include -Wall -g -O2 -pipe -Wall -g -O2 -Wall -Wno-unknown-pragmas -fvisibility=hidden -I/usr/local/include -Wno-long-double -g -O2 -g -O2 -fno-exceptions -fvisibility-inlines-hidden -O3 -fomit-frame-pointer -ffast-math -g2 -fPIC  plugins/video/cursor/cursor.cpp

...skipped cscursor.so for lack of libcrystalspace-1.2.a...
make[1]: *** [all] Error 1
make[1]: Leaving directory `/build/buildd/crystalspace-1.2-20080206/CS'
make: *** [build-arch-stamp] Error 2
dpkg-buildpackage: failure: debian/rules build gave error exit status 2

Please fix this, is really important. The full build log can be found here:

http://launchpadlibrarian.net/15563570/buildlog_ubuntu-intrepid-i386.crystalspace_1.2-20080206-1ubuntu0.3_FAILEDTOBUILD.txt.gz

[vknecht]

Please use latest 1.2.1 stable release, this error is fixed already.

[anonymous]

Already tried with 1.2.1 and it fails to build.

    g++ -Wl,--as-needed -o cscursor.so ./out/linuxx86/optimize/plugins/video/cursor/cursor.o  -Wl,-Bsymbolic-functions -lm -ldl -lnsl -L/usr/local/lib -Wl,-z,defs -Wl,--warn-unresolved-symbols -shared ./out/linuxx86/optimize/libs/libcrystalspace-1.2.a -lpthread -lz -Wl,-Bsymbolic-functions -Wl,-Bsymbolic-functions -lm -ldl -lnsl -L/usr/local/lib -Wl,-z,defs -Wl,--warn-unresolved-symbols \
      -Wl,-soname,cscursor.so
  

  objcopy --add-section .crystalspace=plugins/video/cursor/cscursor.csplugin cscursor.so

make[1]: *** [all] Error 1

[anonymous]

What is the patch that fixes this?

[vknecht]

The output you show us for 1.2.1 doesn't show any actual error,
and certainly not the same as in the first post.
Please find and report the actual error message.
You could for example ask make to stop on first error, if you can't see it.
Also note that Jam is the preferred way to build CS.

[anonymous]

The output doesn't show any error because I used --disable-separate-debug-info flag. How is possible to use Jam to build?

[sunshine]

Replying to anonymous:

What is the patch that fixes this?

The -Wno-long-double issue was resolved in trunk over several revisions as detailed by #463. The changes were applied to the 1.2 branch in r29267.

Replying to vknecht:

Also note that Jam is the preferred way to build CS.

The Makefile just invokes Jam automatically anyhow, so there really is no functional difference. The only noticeable differences are that Makefile enables verbose output and tells Jam to stop after the first error by default (unless make -k is used).

Replying to anonymous:

The output doesn't show any error because I used --disable-separate-debug-info flag. How is possible to use Jam to build?

If you don't have Jam installed separately, you can just invoke ./jam in the CS directory.

Perhaps the error appeared earlier in the output. Given that Makefile is actually invoking Jam, you need to look for the error earlier since Jam dumps error messages first, followed by the command that resulted in the error after.

[anonymous]

Build log for 1.2.1: http://launchpadlibrarian.net/15567604/buildlog_ubuntu-intrepid-i386.crystalspace_1.2.1-0ubuntu0.1_FAILEDTOBUILD.txt.gz

[anonymous]

Some error in the latest snapshot:

        mkdir -p "./out/linuxx86/optimize/plugins/video/cursor"


    g++ -c -o ./out/linuxx86/optimize/plugins/video/cursor/cursor.o -I. -I./include -I./include -Wall -g -O2 -pipe -Wall -g -O2 -Wall -Wno-unknown-pragmas -fvisibility=hidden -I/usr/local/include -g -O2 -g -O2 -fno-exceptions -fvisibility-inlines-hidden -O3 -fomit-frame-pointer -ffast-math -g2 -fPIC  plugins/video/cursor/cursor.cpp

In file included from plugins/video/cursor/cursor.h:23,
                 from plugins/video/cursor/cursor.cpp:38:
./include/ivideo/graph2d.h:146: warning: 'iOffscreenCanvasCallback' is deprecated (declared at ./include/ivideo/graph2d.h:145)

    g++ -Wl,--as-needed -o cscursor.so ./out/linuxx86/optimize/plugins/video/cursor/cursor.o  -Wl,-Bsymbolic-functions -lm -ldl -lnsl -L/usr/local/lib -Wl,-z,defs -Wl,--warn-unresolved-symbols -g2 -shared -Wl,--gc-sections ./out/linuxx86/optimize/libs/libcrystalspace-1.9.so -lz -lpthread -Wl,-Bsymbolic-functions -Wl,-Bsymbolic-functions -lm -ldl -lnsl -L/usr/local/lib -Wl,-z,defs -Wl,--warn-unresolved-symbols -g2 \
      -Wl,-soname,cscursor.so
  

  objcopy --add-section .crystalspace=plugins/video/cursor/cscursor.csplugin cscursor.so


  objcopy --only-keep-debug cscursor.so ./cscursor.dbg
  objcopy --strip-unneeded cscursor.so
  objcopy --add-gnu-debuglink=./cscursor.dbg cscursor.so
  test -n "chmod" && chmod a-x ./cscursor.dbg

make[1]: *** [all] Error 1
make[1]: Leaving directory `/tmp/buildd/crystalspace-1.3-20080624/CS'
make: *** [build-arch-stamp] Error 2
dpkg-buildpackage: failure: debian/rules build gave error exit status 2

[anonymous]

Replying to anonymous:

Some error in the latest snapshot:

Same, not some, sorry.

[sunshine]

Here is the actual error from the build log:

        mkdir -p "./out/linuxx86/optimize/plugins/cscript/csjava/src/org/crystalspace3d"


        mkdir -p "./out/linuxx86/optimize/plugins/cscript/csjava"


    swig -java -c++ -java -package org.crystalspace3d -I./include -I./include -outdir ./out/linuxx86/optimize/plugins/cscript/csjava/src/org/crystalspace3d -o ./out/linuxx86/optimize/plugins/cscript/csjava/csjava.cpp ./include/bindings/cspace.i
    sed 's/\([ 	]*#[ 	]*include[ 	][ 	]*[<"]Python.h[>"]\)/\
    #ifndef __STDC_CONSTANT_MACROS\
    #define __STDC_CONSTANT_MACROS\
    #endif\
    #ifndef __STDC_LIMIT_MACROS\
    #define __STDC_LIMIT_MACROS\
    #endif\
    #ifndef DEBUG_PYTHON\
    #undef _DEBUG\
    #endif\
    \1/g
    /$Header:/d
    s/new[ 	][ 	]*(\([^)][^)]*\))[ 	]*\[/new \1 [/g
    s/[ 	]*#[ 	]*define[ 	][ 	]*SWIG_init[ 	]*init_\(.*\)/#define SWIG_init SWIG_init_\1/g
    s/\([	 ]*#[	 ]*include[	 ][	 ]*[<"]perl.h[>"]\)/\
    #ifndef _DIRENT_H_\
    #define _DIRENT_H_\
    #endif\
    \1/g' < ./out/linuxx86/optimize/plugins/cscript/csjava/csjava.cpp > ./out/linuxx86/optimize/plugins/cscript/csjava/csjava.cpp.sed
    rm -f ./out/linuxx86/optimize/plugins/cscript/csjava/csjava.cpp
    mv -f ./out/linuxx86/optimize/plugins/cscript/csjava/csjava.cpp.sed ./out/linuxx86/optimize/plugins/cscript/csjava/csjava.cpp
  
./include/bindings/java/javapre.i:55: Error: Syntax error in input(1).
/bin/sh: cannot open ./out/linuxx86/optimize/plugins/cscript/csjava/csjava.cpp: No such file
mv: cannot stat `./out/linuxx86/optimize/plugins/cscript/csjava/csjava.cpp.sed': No such file or directory
...skipped <plugins!cscript!csjava!>csjava.o for lack of <plugins!cscript!csjava>csjava.cpp...

This is telling us that there is a syntax error in CS/include/bindings/java/javapre.i at line 55.

That was the only error from the CS portion of that build log (though there were quite a lot of warnings.)

Note that it would be easier to track down these errors if make was told to stop immediately upon the first error (in other words, drop the -k flag from the make invocation).

[sunshine]

Replying to anonymous:

Replying to anonymous:

Some error in the latest snapshot:

Same, not some, sorry.

By snapshot do you mean the trunk (1.3)? If so, you should instruct make to abort at the point where the error occurs so that we can find out exactly what the problem is (and open a separate ticket if it is a different problem). As it stands now, make is continuing after the error, and you are learning that there was a problem only at the end of the make run, but the actual error message is buried somewhere (possibly much) earlier in the build log (not at the end of the make output).

[anonymous]

If I build with --without-java it builds fine. However I don't understand what I should do to provide you more infos about this bug, can you please explain me what I should do?

[sunshine]

Replying to anonymous:

If I build with --without-java it builds fine. However I don't understand what I should do to provide you more infos about this bug, can you please explain me what I should do?

You probably do not need to provide any additional information. Someone needs to examine the syntax error in javapre.i in the 1.2 branch and fix it.

You did, however, mention something about a snapshot, but it was not clear if you were talking about a 1.2.x snapshot or a trunk (1.3) snapshot. If you were talking about the trunk, then we would need to see that error separately and it should be submitted as a distinct bug report, not part of this ticket.

[anonymous]

The bug is in 1.2.1 final version (http://www.crystalspace3d.org/downloads/release/crystalspace-src-1.2.1.tar.bz2) and also in cs-current-snapshot.tar.bz2 tarball (downloaded from http://crystalspace3d.org/cvs-snapshots/bzip2/ , i think is the trunk because it seems to be 1.9 version).

[sunshine]

Replying to anonymous:

The bug is in 1.2.1 final version and also in cs-current-snapshot.tar.bz2 tarball, i think is the trunk because it seems to be 1.9 version).

If the problem is in 1.9, then it also probably is in the new 1.4 branch.

What version of Swig do you have installed?

[anonymous]

swig 1.3.35 as you can see in the build log posted before.

[vknecht]

It would be interesting to check if the errors also occurs with swig 1.3.31 (known to work here) and/or 1.3.33.

[revms]

Replying to vknecht:

It would be interesting to check if the errors also occurs with swig 1.3.31 (known to work here) and/or 1.3.33.

I hit this problem too: 1.3.33 doesn't work. 1.3.31 does, not sure about 32.

[sunshine]

It might be helpful if someone could test with the latest trunk since some fixes were applied in #543. I do not expect that these fixes will resolve this particular issue, but it would be useful to know anyhow.

Further, it would be helpful if someone could tell Swig 1.3.33 to dump the output of its pre-processing step so that we can see the actual error line since the line it reported (55) almost certainly is not the physical line in the input javapre.i file. We will need to look at Swig's pre-processed output to see which line it considers line 55.

To capture the pre-processor output, take the Swig command-line invocation and add the -E option and drop the -o arg option. Then redirect the output to a file, such as swigpostproc.txt. For instance:

swig -java -c++ -java -package org.crystalspace3d -E -I./include -I./include -outdir ./out/linuxx86/optimize/plugins/cscript/csjava/src/org/crystalspace3d ./include/bindings/cspace.i > swigpostproc.txt

Run this command line manually and then attach swigpostproc.txt to this ticket. We also will need to see the actual error message from a normal Swig build run since the line 55 in the 1.2 branch will have moved in the trunk.

[erluk]

With latest trunk (r31355) and Swig 1.3.35, the error is in line 67.

CS/include/bindings/java/javapre.i:67: Error: Syntax error in input(1).

[sunshine]

Replying to erluk:

With latest trunk (r31355) and Swig 1.3.35, the error is in line 67.

Hmm, that line number still does not seem to point at a meaningful source line (containing a syntax error).

[res]

The syntax could be right but the parser broken.

[sunshine]

CS 1.2 build log from original reporter.

[sunshine]

CS 1.2.1 build log from original reporter.

[sunshine]

Replying to erluk:

With latest trunk (r31355) and Swig 1.3.35, the error is in line 67.

I confirm the problem with Swig 1.3.33. Reported line 67 with trunk.

[sunshine]

The line 67 syntax error is resolved by r31357, however the original reported problem is still present.

[sunshine]

Replying to vknecht:

It would be interesting to check if the errors also occurs with swig 1.3.31 (known to work here) and/or 1.3.33.

I just checked, and the syntax error actually is reported even with Swig 1.3.31. The difference, however, with later versions is that Swig 1.3.31 continues processing despite the error. Given that this is flagged as an error (rather than a warning), this should result in an abort, and it is a bug that Swig 1.3.31 does not abort. The behavior of later Swig versions, which do abort, therefore seems correct.

[sunshine]

All of the Swig syntax errors reported in this ticket are resolved by r31360, however, javapre.i still is not free of syntax errors, so further investigation is required.

[sunshine]

Remaining Swig syntax error resolved by r31363.

[sunshine]

Additional linking problem resolved by r31365.

[sunshine]

Also r31371 resolves minor issue with r31365.