On 11/4/20 2:06 am, Michael Westfall wrote:
This time I'm trying to build ncurses-6.2 using the RSB with files .cfg/.bset files I made myself.  Basically, I just copied and modified the files for building libcurl.

Nice. I have used ncurses with RTEMS many many years ago, it is a great package.

Here is what happens when I run the command:
../source-builder/sb-set-builder --prefix=/gem_sw/targetOS/RTEMS/rtems-5.0.0 --host=arm-rtems5 --with-rtems-bsp=beagleboneblack addons/ncurses

The build fails:
rtems-bsp=beagleboneblack addons/ncurses
RTEMS Source Builder - Set Builder, 5.0.0-m2004
Build Set: addons/ncurses
config: addons/ncurses-6.2-1.cfg
package: ncurses-v6.2-arm-rtems5-1
building: ncurses-v6.2-arm-rtems5-1
error: building ncurses-v6.2-arm-rtems5-1
Build FAILED
   See error report: rsb-report-ncurses-v6.2-arm-rtems5-1.txt
error: building ncurses-v6.2-arm-rtems5-1
Build Set: Time 0:00:42.979506
Build FAILED


Checking the error report file, I see:
arm-rtems5-gcc -qrtems -B/gem_sw/targetOS/RTEMS/rtems-5.0.0/arm-rtems5/lib/ -B/gem_sw/targetOS/RTEMS/rtems-5.0.0/arm-rtems5/beagleboneblack/lib/ --specs bsp_specs -mcpu=cortex-a8 -O2 -g -ffunction-sections -fdata-sections  --param max-inline-insns-single=1200  -o blue ../objects/blue.o -L/gem_sw/targetOS/RTEMS/rtems-5.0.0/arm-rtems5/beagleboneblack/lib -mcpu=cortex-a8  -L/gem_sw/targetOS/RTEMS/source/rtems5/rtems-source-builder-5.0.0-m2004/rtems/build/tmp/sb-1000/addons/ncurses/gem_sw/targetOS/RTEMS/rtems-5.0.0/lib -I. -I../../ncurses-6.2/test -I../test -DHAVE_CONFIG_H -DDATA_DIR=\"/gem_sw/targetOS/RTEMS/rtems-5.0.0/arm-rtems5/beagleboneblack/share\" -I. -I../../ncurses-6.2/test -I../include -I../../ncurses-6.2/test/../include -I/gem_sw/targetOS/RTEMS/rtems-5.0.0/arm-rtems5/beagleboneblack/lib/include -D_XOPEN_SOURCE=600 -DNDEBUG -qrtems -B/gem_sw/targetOS/RTEMS/rtems-5.0.0/arm-rtems5/lib/ -B/gem_sw/targetOS/RTEMS/rtems-5.0.0/arm-rtems5/beagleboneblack/lib/ --specs bsp_specs -mcpu=cortex-a8 -O2 -g -ffunction-sections -fdata-sections  --param max-inline-insns-single=1200  `echo "-L../lib -lform -lmenu -lpanel -lncurses   -lbsd -lm -lz -lrtemsdefaultconfig" | sed -e 's/-lform.*-lpanel[^ ]*//'` arm-rtems5-gcc: fatal error: /gem_sw/targetOS/RTEMS/rtems-5.0.0/arm-rtems5/beagleboneblack/lib/bsp_specs: attempt to rename spec 'endfile' to already defined spec 'old_endfile'
compilation terminated.
Makefile:270: recipe for target 'background' failed
make[1]: *** [background] Error 1

The problem, I'm guessing is that --specs bsp_specs is included twice in the compiler command line.

<sigh>

What am I missing to stop this from happening?
I don't see anything the RSB documentation to help me with this.

I would add `--trace` and then look in the ncurses' configure generated config.log file to see how the linker command line is created. I would look at CFLAGS, LDFLAGS, CCFLAGS or one of the other types of flags to see how they are being used. The other possible option is to see if you can avoid examples or commands being linked?

An aside, Joel has done a lot of background work to remove spec files from our compiler command line and Sebastian is pushing patches looking to removing spec files from RTEMS 6. Lets hope this is successful as this type of error disappears. There benefits of having a "normal" compiler command line cannot be stated enough.

I've attached my .cfg and .bset files.

They look good. I suspect something in the default pre-build set up is effecting something ...

https://git.rtems.org/rtems-source-builder/tree/source-builder/defaults.mc#n212

This macro string is expanded and written into the script file `do-build` so you can edit the generated script file and run it to see what effect it has. In summary debugging this is a matter of seeing how the pre-build variables are effecting the flags in the package and config.log and the generated makefile will hopefully provide some clues. You can then look at overriding the settings in your cfg file.

Chris
_______________________________________________
users mailing list
[email protected]
http://lists.rtems.org/mailman/listinfo/users

Reply via email to