Bugzilla Bug 130023

Someone make this work.  Gives me sandbox violations from scons.

Created an attachment (id=84686) [edit]
dangerdeep-0.1.1.ebuild

Not so working ebuild.

*** Bug 130191 has been marked as a duplicate of this bug. ***

Created an attachment (id=84851) [edit]
games-simulation/dangerdeep-0.1.1-r1.ebuild

This new ebuild works perfectly on amd64. Can someone test it on x86?

Created an attachment (id=84852) [edit]
files/dangerdeep-destdirs.patch

(From update of attachment 84851 [edit])
-r is for in-portage rev bumps

Should dynamically replace the directories with the variables from games.eclass
(GAMES_BINDIR, GAMES_DATADIR...)

Also, after reading http://ponto-dot.com/dftd/viewtopic.php?t=42, I suggest the
following src_compile() :

    if use amd64
    then
        USEX86SSE=3
    else
        if use sse
        then
            USEX86SSE=1
        else
            USEX86SSE=-1
        fi
    fi

    scons usex86sse=${USEX86SSE} || die "make failed"

and adding sse to the USE flags.

This will use asm sse optimizations on x86 when sse flag is set, and intrisics
ones by default for amd64 (since all amd64 have sse). According to the post,
the performance boost is quite nice ;)

I built it on amd64. I noticed that

- dangerdeep depends on media-libs/libsdl built with joystick
- I got some access violations caused by scons

I will post the actual ACCESS_VIOLATION when an update is there.

Other than that it builts and runs fine.

Created an attachment (id=85635) [edit]
dangerdeep-0.1.1.ebuild

Created an attachment (id=85636) [edit]
dangerdeep-0.1.1-build.patch

Respect flags and fix sandbox violations

I just tried the new ebuild but it failed with these errors:

 * Applying dangerdeep-0.1.1-build.patch ...                                   
                                                      [ ok ]
>>> Source unpacked.
>>> Compiling source in /var/tmp/portage/dangerdeep-0.1.1/work/dangerdeep-0.1.1 ...
scons: Reading SConscript files ...
Compiling for Unix/Posix/Linux Environment
KeyError: 'LDFLAGS':
  File "SConstruct", line 180:
    env.Append(LINKFLAGS = os.environ['LDFLAGS'])
  File "/usr/lib64/python2.4/UserDict.py", line 17:
    def __getitem__(self, key): return self.data[key]

!!! ERROR: games-simulation/dangerdeep-0.1.1 failed.
Call stack:
  ebuild.sh, line 1525:   Called dyn_compile
  ebuild.sh, line 928:   Called src_compile
  dangerdeep-0.1.1.ebuild, line 43:   Called die

!!! scons failed


I am running amd64 unstable.

Ok, I should have had a closer look.
I didn't have any LDFLAGS specified in my make.conf and that's why it failed.
So I added LDFLAGS in it and now it compiles.

Created an attachment (id=85730) [edit]
dangerdeep-0.1.1-build.patch

You shouldn't have to have it set. Try this one...

It seems to work now, with and without settings the LDFLAGS.

Thank you :)

Created an attachment (id=85740) [edit]
dangerdeep-0.1.1-build.patch

This is probably better. Sorry for spam.

segfaults for me.

Created an attachment (id=94911) [edit]
dangerdeep-0.1.1.ebuild

This doesn't really depend on boost.

Created an attachment (id=94912) [edit]
dangerdeep-0.1.1-build.patch

Use the right compiler.

*** Bug 110121 has been marked as a duplicate of this bug. ***

Created an attachment (id=99472) [edit]
New ebuild

Version 0.2.0 ebuild only tested on ~amd64.
My first try at ebuilds so be gentle ;)

addwrite /usr/include/GL:/usr/include/SDL:/usr/X11R6/lib is pretty evil.  Feel
like tracking down why that's needed and fixing it?

Still cores for me right after trying to start a single mission:

0x806ce43 in dangerdeep at ??:0
0x804ebe2 in dangerdeep at ??:0
0xffffe420 in [0xffffe420] at ??:0
0xb7cfa876 in /usr/lib/libfftw3f.so.3 at ??:0
0xb7cfb634 in fftwf_mkplan_dft_ct at ??:0
0xb7cfa9f5 in /usr/lib/libfftw3f.so.3 at ??:0
0xb7cf623d in /usr/lib/libfftw3f.so.3 at ??:0
0xb7cf6b9c in fftwf_mkplan_d at ??:0
0xb7cfe765 in /usr/lib/libfftw3f.so.3 at ??:0
0xb7cf623d in /usr/lib/libfftw3f.so.3 at ??:0
0xb7cf6b9c in fftwf_mkplan_d at ??:0
0xb7cfe765 in /usr/lib/libfftw3f.so.3 at ??:0
0xb7cf623d in /usr/lib/libfftw3f.so.3 at ??:0
0xb7cf6b9c in fftwf_mkplan_d at ??:0
0xb7cfb679 in fftwf_mkplan_dft_ct at ??:0
0xb7cfb425 in /usr/lib/libfftw3f.so.3 at ??:0

That's fftw-3.0.1-r2

hrm, I'm using sci-libs/fftw-3.1.2, the addwrite is for scons, I'm not really
into python and such but I know it plays silly buggers with the sandbox and
that was the only way I could get it to work. If anyone knows a better way
around it ?

The build patch from 0.1.1 ebuild solves the sandbox problems, adding it back
to 0.2.0 worked here. 
I also changed the logic for SSE use, since sse use flag is always filtered on
amd64 ans so cannot be enabled

No segfaults here either with unstable fftw, is one of them ready for
stabilizing to solve the problem?

Created an attachment (id=99508) [edit]
Updated 0.2.0 ebuild

* Removed addwrite
* Added back build patch
* Changed sse logic

Created an attachment (id=99509) [edit]
Build patch updated for 0.2.0

Above patch and ebuild work perfectly for me on ~amd64

Created an attachment (id=99829) [edit]
dangerdeep-artwork license

The license that the artwork is distributed under, it is included in the data
zip, put here for easy of access.

The ebuild is working perfectly here.

Please put it in portage..

In portage.