Compile for arm-linux-gnueabihf? about koreader-base HOT 8 CLOSED

I've solved this, it required a little more than a "make clean" since it didn't actually remove all the cruft. In the end I also changed to ARM_ARCH:=-march=armv6j -mtune=arm1136jf-s -mfpu=vfp -mfloat-abi=hard -marm

from koreader-base.

oh, cool, do we also need to adapt the makefile?

from koreader-base.

Hm. Shouldn't specifying CHOST on the make command line be enough? like e.g.

$ make CHOST=arm-linux-gnueabihf

That way, all subsequent definitions via ?= and := should be overridden, I
think. Or was it via environment variable (like in "CHOST=... make")? I'm
pretty sure I switched between compilers that way pretty easily.

BTW, we should identify what cruft is left by "make clean" - as that may
give us strange bugshere and there when not working properly.

from koreader-base.

It looked like thirdparty libs and luajit were not properly cleaned. Yes, specifying CHOST looks to be enough, but I don't think you can change to -mfloat-abi=hard by a command line switch, right?

from koreader-base.

Yes, that may pose some problems, I guess. I'm not too sure, though. Isn't
that the default value for a hf-toolchain, though?

from koreader-base.

it is but in Makefile.defs it is specified:
ARM_ARCH:=-march=armv6j -mtune=arm1136jf-s -mfpu=vfp -mfloat-abi=softfp -marm
so if you don't change it in the end you get a strange hybrid I guess. I didn't try though to leave it unchanged and see if the produced binary would run.

from koreader-base.

Oh, right. It probably won't run, no - unsure if it would even link...

from koreader-base.

yeah, we can change ARM_ARCH according to environment. So if you think it's necessary, we should do so.

btw, what command did you run to clean? make clean in koreader or koreader-base?

from koreader-base.