tests/lib/mem_alloc newlibnano target run time error

[hakehuang]

Describe the bug
ASSERTION FAIL [(options & ((1UL << (2)))) == 0] @ ZEPHYR_BASE/kernel/thread.c:581
Platform is capable of user mode, and test thread created with K_USER option, but neither CONFIG_TEST_USERSPACE nor CONFIG_USERSPACE is set

To Reproduce
Steps to reproduce the behavior:

1. mkdir build; cd build
2. cmake -DBOARD=frdm_k64f ..
3. make
4. make flash
5. see log

Expected behavior
runs sucessfully

Impact
test coverage

Screenshots or console output

***** Booting Zephyr OS version 2.1.0-rc1 *****
Running test suite test_c_lib_dynamic_memalloc
===================================================================
starting test - test_malloc
ASSERTION FAIL [(options & ((1UL << (2)))) == 0] @ ZEPHYR_BASE/kernel/thread.c:581
       Platform is capable of user mode, and test thread created with K_USER option, but neither CONFIG_TEST_USERSPACE nor CONFIG_USERSPACE is set
E: r0/a1:  0x00000004  r1/a2:  0x00000245  r2/a3:  0x00800000
E: r3/a4:  0x600095d2 r12/ip:  0x150b0000 r14/lr:  0x60006cd5
E:  xpsr:  0x61000000
E: Faulting instruction address (r15/pc): 0x60007732
E: >>> ZEPHYR FATAL ERROR 4: Kernel panic on CPU 0
E: Current thread: 0x2000017c (main)
E: Halting system

Environment (please complete the following information):

• OS: (Linux)
• Toolchain (Zephyr SDK, ...)
• Commit SHA or Version used: v2.1.0-rc1

[hakehuang]

@dleach02

[stephanosio]

It looks like 7b1ee5c missed prj_newlibnano.conf (probably because it is not run when using Zephyr SDK due to toolchain_whitelist: gnuarmemb).

Adding CONFIG_TEST_USERSPACE=y to prj_newlibnano.conf should resolve this issue.

p.s. do we really want CONFIG_TEST_USERSPACE when newlibnano is used? that seems to go against what the commit referenced above is trying to solve, in which case we should not be using ztest_user_unit_test. cc @andrewboie

[ioannisg]

there

It looks like 7b1ee5c missed prj_newlibnano.conf (probably because it is not run when using Zephyr SDK due to toolchain_whitelist: gnuarmemb).

Adding CONFIG_TEST_USERSPACE=y to prj_newlibnano.conf should resolve this issue.

There is some inconsistency here: the test-case for newlibnano is tagged with -userspace, but the prj_newlibnano.conf does not enable it.

Agree that CI does not catch this because it runs with zephyr toolchain...

p.s. do we really want CONFIG_TEST_USERSPACE when newlibnano is used? that seems to go against what the commit referenced above is trying to solve, in which case we should not be using ztest_user_unit_test. cc @andrewboie

What the commit does is to require TEST_USERSPACE be enabled, if the test code defines User threads, right? So yeah, either we stop using ztest_user_unit_test if !USERSPACE in main.c or we add TEST_USERSPACE for newlibnano..

[andrewboie]

If the test does stuff in user mode, CONFIG_TEST_USERSPACE needs to be enabled.

[ioannisg]

If the test does stuff in user mode, CONFIG_TEST_USERSPACE needs to be enabled.

Well, the tests is testing dynamic memory allocation, and that's coupled with user space. We should therefore run the test in user mode, imho.