Problem building the SDK in Cygwin.

[Dr.BeauWebber]

Hi,
I have been trying for a few days to build the SDK in Cygwin. (I have been using Cygwin for many years).
I believe I have all the pre-requisites installed, no error messages.

My first problem is that the build shell seems to at some stage hang -
perhaps this is related to when I leave the lap-top running overnight, but the power setting is never to sleep.

4 bld-epiphany/
1088 build-2013-05-21-2046.log
244 build-2013-05-22-0923.log
432 build-2013-05-22-1517.log
1024 build-2013-05-22-2243.log
884 build-2013-05-23-1310.log
16 gcc/
0 INSTALL/
4 sdk/
16 src/
20 srcw/


Today the machine certainly never slept, but still I received a fail message :
Creating unified source...
Configuring tools...
Building tools...
Error: Build failed.
**************************************************
The SDK build has failed.

In the script I see at line 4868 :
checking for struct reg.r_gs... failed

and at the end :

ar rc libsim.a callback.o syscall.o targ-map.o version.o sim-arange.o sim-bits.o sim-command.o sim-config.o sim-core.o sim-endian.o sim-events.o sim-fpu.o sim-io.o sim-info.o sim-load.o sim-memopt.o sim-module.o sim-options.o sim-profile.o sim-signal.o sim-trace.o sim-utils.o sim-watch.o sim-cpu.o sim-hload.o sim-hrw.o sim-model.o sim-reg.o cgen-utils.o cgen-trace.o cgen-scache.o cgen-fpu.o cgen-accfp.o cgen-run.o sim-reason.o sim-engine.o sim-stop.o sim-if.o arch.o epiphany.o cpu.o decode.o sem.o model.o mloop.o epiphany-fp.o devices.o traps.o
rm gfdl.pod cpp.pod gcov.pod gcc.pod fsf-funding.pod
make[1]: Leaving directory `/cygdrive/d/home/jbww/Src/Git/Parallella/bld-epiphany/gcc'
Makefile:5800: recipe for target `all-gcc' failed
make: *** [all-gcc] Error 2
make: *** Waiting for unfinished jobs....
ranlib libsim.a
gcc -DHAVE_CONFIG_H -DWITH_DEFAULT_MODEL='"epiphany32"' -DPROFILE=1 -DWITH_PROFILE=-1 -DWITH_ALIGNMENT=STRICT_ALIGNMENT -DWITH_TARGET_BYTE_ORDER=LITTLE_ENDIAN -DWITH_ENVIRONMENT=ALL_ENVIRONMENT -DWITH_HOST_BYTE_ORDER=LITTLE_ENDIAN -DWITH_SCACHE=16384 -Wimplicit -Wreturn-type -Wcomment -Wtrigraphs -Wformat -Wparentheses -Wpointer-arith -Wuninitialized -I. -I/cygdrive/d/home/jbww/Src/Git/Parallella/srcw/sim/epiphany -I../common -I/cygdrive/d/home/jbww/Src/Git/Parallella/srcw/sim/epiphany/../common -I../../include -I/cygdrive/d/home/jbww/Src/Git/Parallella/srcw/sim/epiphany/../../include -I../../bfd -I/cygdrive/d/home/jbww/Src/Git/Parallella/srcw/sim/epiphany/../../bfd -I../../opcodes -I/cygdrive/d/home/jbww/Src/Git/Parallella/srcw/sim/epiphany/../../opcodes -g -O2 -Wl,--stack,12582912 -o run.exe \
nrun.o libsim.a ../../bfd/libbfd.a ../../opcodes/libopcodes.a -lintl ../../libiberty/libiberty.a -lz -lm
make[2]: Leaving directory `/cygdrive/d/home/jbww/Src/Git/Parallella/bld-epiphany/sim/epiphany'
make[2]: Entering directory `/cygdrive/d/home/jbww/Src/Git/Parallella/bld-epiphany/sim/common'
make[2]: Nothing to be done for `default'.
make[2]: Leaving directory `/cygdrive/d/home/jbww/Src/Git/Parallella/bld-epiphany/sim/common'
make[1]: Leaving directory `/cygdrive/d/home/jbww/Src/Git/Parallella/bld-epiphany/sim'

Any suggestions as to where the problem may lie ?
cheers,
Beau

[Hoernchen]

You need to post/pastebin a few more lines of your log file, the actual error is not in the part you posted.

[simoncook]

Hi,

As Hoernchen has said, could you place somewhere the entire log file for us to look at. As the build occurs in parallel, the place where the error occurs isn't necessary right at the bottom of the log as the output of other parts being built may appear underneath.

[Dr.BeauWebber]

Thanks,
I have had a number of attempts to build the SDK, and it seems to get a different distance each time :
$ ls -FCs1
total 5914
0 bld-epiphany_2/
4 bld-epiphany_3/
4 bld-epiphany_4/
4 bld-epiphany_5/
1088 build-2013-05-21-2046.log
244 build-2013-05-22-0923.log
432 build-2013-05-22-1517.log
980 build-2013-05-22-2243.log
884 build-2013-05-23-1310.log
1064 build-2013-05-23-2000.log
1 build-2013-05-24-0051.log
836 build-2013-05-24-0940.log
244 build-2013-05-25-1020.log
1 build-2013-05-25-1536.log
16 gcc/
0 INSTALL_2/
0 INSTALL_3/
0 INSTALL_4/
0 INSTALL_5/
4 sdk/
16 src/
4 srcw/
8 srcw_1/
20 srcw_2/
20 srcw_3/
20 srcw_4/
20 srcw_5/

A couple of these have ended when I have had to hibernate the lap-top to transport it, others I do not know why.
I attach the last log.
I am trying once more.
cheers,
Beau

[Dr.BeauWebber]

Another failure.
Log attached.

Ahh, this time both have failed at :
Makefile:760: recipe for target `chew.exe' failed

cheers,

[Hoernchen]

There is something weird going on, I did not have any problems like that when I built parts of the sdk (binutils, gcc) using mingw on windows. Are you using -jX with X >1 ?

[Dr.BeauWebber]

Hmm,
I think I may have a broken machine.
I have been working on the assumption that some piece of software (such as Norton) is waking up and shutting the compilation down, but I have turned off the internet and shut Norton down as far as I can.
It then runs faster, but still dies.

Note sure about that flag : I am using the build script : build-toolchain.sh
as the docs suggest :
This has :
# If /proc/cpuinfo is avaliable, limit load to launch extra jobs to
# number of processors + 1, otherwise use a constant of 2.
make_load="-j -l `(echo processor;cat /proc/cpuinfo 2>/dev/null || echo processor)|grep -c processor`"

and yes, multiple processes are running, 8 processors mostly near 100%.

I have switched off rebuilding the sources each time, but sometimes it ends in a failure message and sometimes it still just hangs.
I am now trying on a very old 2 processor XP machine.

[timpart]

I have successfully built the Embecosm SDK on windows 7 cygwin. Make sure you have the latest version of automake installed. Older versions weren't generating the right C code from the chip configuration file. This doesn't seem to be related to the problem you are having but might save time later. If you have fixed problems already building from scratch might help, but clearly will take a lot of time.

Tim

[Dr.BeauWebber]

Thanks for that - I started by updating Cygwin, when I installed Git, so hopefully that will be OK.
cheers, beau

[Dr.BeauWebber]

OK, I have succeeded in building the tool-chain, and e-gcc and e-run successfully build and run helloworld.c

OK, I believe I have a problem with my laptop that only shows with very long compilations, but I am there now :

Now clearly, if the source files are updated, you need to rebuild each section from scratch, but what I have done is
first build the sources, then comment that section out from a copy of the build-toolchain.sh script,
and then make use of make's property of not re-doing work unnecessary by not deleting the work already dome, but after a machine hang, just re-running the build.
On the third iteration I now have a working gcc tool-chain in Cygwin (but no IDE).
cheers,
Beau