[Ohrrpgce] SVN: james/12141 Docker: add a first-draft attempt at an mxe build environment in a docke

Sat Dec 26 13:33:01 PST 2020

Okay! Thank you for all those good clarifications.

I tried to incorporate them into my Dockerfile, but no luck yet. Right now
it is failing with

CC src/rtlib/obj/linux-x86/io_print.o
CC src/rtlib/obj/linux-x86/io_print_bool.o
src/rtlib/linux/io_mouse.c:17:17: fatal error: gpm.h: No such file or
directory
 #include <gpm.h>
                 ^
compilation terminated.
makefile:670: recipe for target 'src/rtlib/obj/linux-x86/io_mouse.o' failed
make: *** [src/rtlib/obj/linux-x86/io_mouse.o] Error 1
make: *** Waiting for unfinished jobs....
The command '/bin/sh -c make compiler install-compiler install-includes -j6
    && make rtlib install-rtlib gfxlib2 install-gfxlib2 MULTILIB=32 -j6
&& make rtlib install-rtlib gfxlib2 install-gfxlib2 MULTILIB=64 -j6     &&
make unit-tests     && echo "Done make linux and unit tests..."' returned a
non-zero code: 2

On Wed, Dec 23, 2020 at 7:31 PM Ralph Versteegen <teeemcee at gmail.com> wrote:

> It isn't necessary to use a config.mk file. I ran
> make rtlib install-rtlib gfxlib2 install-gfxlib2
> TARGET=i686-w64-mingw32.static prefix=$HOME/local/fbc-git CFLAGS="-g
> -DDISABLE_FFI"
> and it worked for me. This should have worked regardless of which fbc
> branch you were on; I didn't add that flag.
>
> Then I had a look at mxe's libffi test program and found that they use
> pkg-config to get the include path. The following works:
> TARGET=i686-w64-mingw32.static
> make rtlib install-rtlib gfxlib2 install-gfxlib2 TARGET=$TARGET
> prefix=$HOME/local/fbc-git CFLAGS="-g `$TARGET-pkg-config libffi --cflags`"
> I'm guessing that ffi.h is hidden like this because it's arch-specific,
> unlike normal headers.
>
> On Thu, 24 Dec 2020 at 03:54, Ralph Versteegen <teeemcee at gmail.com> wrote:
>
>>
>>
>> On Thu, 24 Dec 2020 at 03:52, Ralph Versteegen <teeemcee at gmail.com>
>> wrote:
>>
>>>
>>>
>>> On Thu, 24 Dec 2020 at 03:45, Ralph Versteegen <teeemcee at gmail.com>
>>> wrote:
>>>
>>>> Oh, sorry, I did write up instructions for compiling but didn't send
>>>> them yet.
>>>>
>>>> In order to compile a fbc that can target android you of course need to
>>>> use https://github.com/rversteegen/fbc.git which I have just
>>>>
>>>
>> Argh, when I copy-pasted that link from github I failed to realise it
>> didn't include the branch name. I meant my android branch from there.
>>
>>
>>> merged upstream master into (and which therefore support emscripten
>>>> too!) However this branch doesn't have the mac branch merged into it. That
>>>> doesn't matter until we have a cross-compiler that targets Mac. At that
>>>> point I can merge them. (Please ignore my androidmac branch, it's very old
>>>> and I can't even remember what's going on in it.
>>>>
>>>> Anyway, to target android it isn't necessary to build the rtlib or
>>>> gfxlib2 for it, unlike other platforms, due to our hacky android build
>>>> system.
>>>>
>>>
>>> To clarify: I used my android branch and compiled linux 32/64-bit and
>>> windows libraries using it to get a multi-target FB installation.
>>>
>>> Also, I forgot to mention that recently when I tried to compile
>>> ohrrpgce-game for Android I found it would crash when entering BROWSE(). I
>>> didn't figure out why yet. I wonder whether it's due to something that went
>>> wrong when I last merged upstream master into the android branch, back in
>>> June. I think I must have successfully compiled for Android more recently
>>> than that though. Would like to hear whether it works for you. I haven't
>>> tried again since doing the latest merge.
>>>
>>>
>>>> #!/bin/sh
>>>>
>>>> cat <<EOF > config.mk
>>>> prefix=$HOME/local/fbc-git
>>>> CFLAGS += -Wfatal-errors -g
>>>> #FBC := ~/local/fbc-1.05/bin/fbc
>>>>
>>>> # For some reason mxe puts ffi.h in a separate include dir, so it can't
>>>> be found.
>>>> # libffi is only needed for FB's ThreadCall operator, which we don't
>>>> use, so just disable it.
>>>> CFLAGS += -DDISABLE_FFI
>>>>
>>>> EOF
>>>>
>>>> # Compiler and includes
>>>> make compiler install-compiler install-includes -j6  || exit 1
>>>> # Linux x86 libraries
>>>> make rtlib install-rtlib gfxlib2 install-gfxlib2 MULTILIB=32 -j6  ||
>>>> exit 1
>>>> # Linux x86_64
>>>> make rtlib install-rtlib gfxlib2 install-gfxlib2 MULTILIB=64 -j6  ||
>>>> exit 1
>>>> make unit-tests  || exit 1
>>>> # Note: make doesn't return false if any log-tests fail. A number do
>>>> fail,
>>>> # should doublecheck none are due to changes on the android branch.
>>>> #make log-tests  || exit 1
>>>> # Windows x86 (using mxe; mxe's usr/bin must be in PATH)
>>>> make rtlib install-rtlib gfxlib2 install-gfxlib2
>>>> TARGET=i686-w64-mingw32.static -j6  || exit 1
>>>>
>>>> # The following are for compiling commandline Android test programs,
>>>> but NOT needed for Game/Custom builds.
>>>> # export PATH=$HOME/local/android-toolchain-r12b-api9-arm/bin:$PATH
>>>> # export PATH=$HOME/local/android-toolchain-r12b-api17-x86/bin:$PATH
>>>> # make rtlib install-rtlib TARGET=arm-linux-androideabi -j6  || exit 1
>>>> # make rtlib install-rtlib TARGET=i686-linux-android -j6  || exit 1
>>>>
>>>>
>>>>
>>>> On Wed, 23 Dec 2020 at 09:38, James Paige <Bob at hamsterrepublic.com>
>>>> wrote:
>>>>
>>>>> Right now this always fails in the step that does:
>>>>>
>>>>> RUN make TARGET=i686-w64-mingw32.static prefix=/usr/local/ rtlib
>>>>> gfxlib2 install-rtlib install-gfxlib2
>>>>>
>>>>> It fails with this error message about ffi.h
>>>>>
>>>>> CC src/rtlib/obj/win32/sys_portio.o
>>>>> CC src/rtlib/obj/win32/sys_rmdir.o
>>>>> CC src/rtlib/obj/win32/sys_run.o
>>>>> CC src/rtlib/obj/win32/sys_shell.o
>>>>> CC src/rtlib/obj/win32/thread_call.o
>>>>> src/rtlib/thread_call.c:33:17: fatal error: ffi.h: No such file or
>>>>> directory
>>>>> compilation terminated.
>>>>> makefile:554: recipe for target 'src/rtlib/obj/win32/thread_call.o'
>>>>> failed
>>>>> make: *** [src/rtlib/obj/win32/thread_call.o] Error 1
>>>>> The command '/bin/sh -c make TARGET=i686-w64-mingw32.static
>>>>> prefix=/usr/local/ rtlib gfxlib2 install-rtlib install-gfxlib2' returned a
>>>>> non-zero code: 2
>>>>>
>>>>> libffi-dev is definitely installed, so i don't know what the problem
>>>>> is.
>>>>>
>>>>> I tried skipping libffi with CFLAGS=-DDISABLE_FFI per the instructions
>>>>> in https://rpg.hamsterrepublic.com/ohrrpgce/Cross-compiling but that
>>>>> made no difference.
>>>>>
>>>>> I also tried switching from https://github.com/freebasic/fbc.git to
>>>>> https://github.com/rversteegen/fbc for the FBC sources, and that
>>>>> didn't help either (unless maybe I needed to use a specific branch?)
>>>>>
>>>>> Anyway, I'll look at it more later. This was as far as I got today :D
>>>>>
>>>>> ---
>>>>> James Paige
>>>>>
>>>>> On Tue, Dec 22, 2020 at 3:32 PM <subversion at hamsterrepublic.com>
>>>>> wrote:
>>>>>
>>>>>> james
>>>>>> 2020-12-22 12:32:41 -0800 (Tue, 22 Dec 2020)
>>>>>> 158
>>>>>> Docker: add a first-draft attempt at an mxe build environment in a
>>>>>> docker container.
>>>>>>
>>>>>> This does not work yet, but will provide a base for further
>>>>>> improvements
>>>>>> ---
>>>>>> A   wip/docker/ohrrpgce-mxe-build-env/
>>>>>> A   wip/docker/ohrrpgce-mxe-build-env/Dockerfile
>>>>>>
>>>>>> _______________________________________________
>>>>>> Ohrrpgce mailing list
>>>>>> ohrrpgce at lists.motherhamster.org
>>>>>> http://lists.motherhamster.org/listinfo.cgi/ohrrpgce-motherhamster.org
>>>>>>
>>>>> _______________________________________________
>>>>> Ohrrpgce mailing list
>>>>> ohrrpgce at lists.motherhamster.org
>>>>> http://lists.motherhamster.org/listinfo.cgi/ohrrpgce-motherhamster.org
>>>>>
>>>> _______________________________________________
> Ohrrpgce mailing list
> ohrrpgce at lists.motherhamster.org
> http://lists.motherhamster.org/listinfo.cgi/ohrrpgce-motherhamster.org
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.motherhamster.org/pipermail/ohrrpgce-motherhamster.org/attachments/20201226/6b7c7cac/attachment-0001.html>