shell/sc_can: fix uninitialized warning

[cladmi]

Contribution description

When compiled for hifive1 board with gcc-7.2.0 this warning was raised:

'ret' may be used uninitialized in this function [-Werror=maybe-uninitialized]

This was first mentioned in RIOT-OS/riotdocker#42 (comment)

Testing

The board does not have enough memory but the error is triggered when compiling with a gcc version 7.2.0:

make -C tests/conn_can/ BOARD=hifive1 RIOT_CI_BUILD=1

Warning is dismissed, it was just a testing procedure error.

Warning

Currently when compiling with riscv-none-embed-gcc (GNU MCU Eclipse RISC-V Embedded GCC, 64-bits) 7.2.0 a linker error is raised

make -C tests/conn_can/ BOARD=hifive1
...
/opt/gnu-mcu-eclipse/riscv-none-gcc/7.2.0-2-20180111-2230/bin/../lib/gcc/riscv-none-embed/7.2.0/../../../../riscv-none-embed/bin/ld: section .stack VMA [0000000080003c00,0000000080003fff] overlaps section .bss VMA [00000000800001f8,0000000080003eaf]
collect2: error: ld returned 1 exit status
/data/riotbuild/riotbase/Makefile.include:356: recipe for target '/data/riotbuild/riotproject/tests/conn_can/bin/hifive1/tests_conn_can.elf' failed
make: *** [/data/riotbuild/riotproject/tests/conn_can/bin/hifive1/tests_conn_can.elf] Error 1
/home/user/Desktop/RIOT_master_old/makefiles/docker.inc.mk:90: recipe for target '..in-docker-container' failed
make: *** [..in-docker-container] Error 2

The goal of this PR is to resolve the linking bug and merge this.

Issues/PRs references

#9405
RIOT-OS/riotdocker#42
Found it during release testing

[cladmi]

@kenrabold Could you look if you have the same issue with linking in this PR ?

I raise the warning when doing make -C tests/conn_can/ BOARD=hifive1

[cladmi]

I add a waiting for other PR because it needs a fix for the linking error before being merged.

[kenrabold]

The link failure is from insufficient memory on the hifive1 board to hold the program image and RAM sections. The hifive1 board is already in the BOARD_INSUFFICIENT_MEMORY list for the /tests/conn_can Makefile, so this test app should never be compiled for hifive1.

[cladmi]

@kenrabold Thank you for your quick feedback, and sorry for not paying enough attention.

The message was different from what I see in arm boards arm-none-eabi/bin/ld: region 'ram' overflowed by XXXbytes so I thought it could was something else.

Also I was not being careful enough when testing.
I had an error when doing buildtest, that only builds without linking, and after fixing the warning that is here, I tested with make all instead of buildtest or make BOARD=hifive1 RIOT_CI_BUILD=1 which tries to link and does not consider BOARD_INSUFFICIENT_MEMORY, which mislead me.

So in fact no problems with hifive1 board, I am relieved.

[ZetaR60]

ACK. I made the same change in #9507 to get hifive1 working properly, but it is probably better as a separate PR. I have done build tests with this change in #9507.

[cladmi]

@ZetaR60 Yes I like fixing what is possible in sub-PRs as it helps getting merged early. Also it has been problematic for me in the release tests.

[cladmi]

Backport provided in #9644