Fix array vs pointer error in declaration of circuitpython_help_text

[jepler]

.. removing the need for -Wno-error=lto-type-mismatch, which is
not available in gcc 5.4.1 (debian stretch) while building the atmel-samd port.

The specific error that occurred was:

../../py/builtin.h:121:19: error: type of 'circuitpython_help_text' does not match original declaration [-Werror]
 extern const char MICROPY_PY_BUILTINS_HELP_TEXT[];
                   ^
../../shared-bindings/help.c:38:13: note: previously declared here
 const char *circuitpython_help_text =
             ^
lto1: all warnings being treated as errors
lto-wrapper: fatal error: /usr/bin/arm-none-eabi-gcc returned 1 exit status
compilation terminated.
/usr/lib/gcc/arm-none-eabi/5.4.1/../../../arm-none-eabi/bin/ld: error: lto-wrapper failed
collect2: error: ld returned 1 exit status

[dhalbert]

@jepler We build with gcc 7.3.1 from https://developer.arm.com/open-source/gnu-toolchain/gnu-rm/downloads (7-2017-q4-major) or https://launchpad.net/~team-gcc-arm-embedded/+archive/ubuntu/ppa (if using Ubuntu). I'd recommend that you download that toolchain and put it in user space or use Ubuntu.

We need the -Wno-error=lto-type-mismatch to fix an issue when building frozen modules with gcc 7.3.1. This issue comes up when doing the BOARD=circuit_playground_express. You can see the issue in the travis builds for that board. So unfortunately this PR will not work for us.

[jepler]

Thanks, I see that failure on travis-ci now. Oddly, I don't replicate the mp_qstr_const_pool problem here, with make -C ports/atmel-samd/ BOARD=circuitplayground_express.

Do you think there's any point in revising this pull request? If not, please feel free to close it up.

Do you think there's any point in opening a documentation pull request? I do not find these toolchain requirements listed in the docs, even after grepping for the URLs you cited above. (some of the other ports/*/README.md do mention specific gcc requirements, but not ports/atmel-samd/README.md)

[jepler]

For instance, a revised patch might use the Linux kernel trick for checking whether a particular argument is accepted by the compiler or not, similar to

cc-option = $(shell if $(CC) $(CFLAGS) $(1) -S -o /dev/null -xc /dev/null \
             > /dev/null 2>&1; then echo "$(1)"; else echo "$(2)"; fi ;)
...
CFLAGS += $(call cc-option,-mieee-fp)

[dhalbert]

I agree we should list the toolchain requirements. That is an oversight: thanks for pointing that out. A PR to document that would be great, if you are willing, or we can just do it in a future PR of ours.

I submitted a gcc bug report for the bogus LTO mismatch warning. It is now fixed in gcc, but will not show up for a while: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81440.

You may also run into other issues with the older gcc, such as code size. Right now we have a few kbytes free on the non-Express builds, but we may try squeezing more stuff in later. At one time we only had a few tens or hundreds of bytes free, so the better optimization on the newer compilers was very helpful.

[jepler]

Thanks for the pointer to the upstream bug.

I've revised THIS pull request so that it only changes the declaration, and does NOT modify the compiler flags.

[tannewt]

Thanks for the tweak! Glad to see more people building CircuitPython locally.