testing/ostest: Exercise custom work queues.

[13022591351]

Summary

• Extend the existing ostest work queue coverage to dynamically created
  queues with one and two workers.
• Run scheduler queues and custom queues in Flat builds, and libc custom user
  queues in non-Flat builds.
• Add ostest wqueue as a focused entry without changing the original
  no-argument ostest flow.
• Cover explicit caller priorities, invalid arguments, periodic requeue,
  pending replacement, synchronous and parallel cancellation, multiple
  simultaneous queues, self-destruction rejection, and teardown.
• Check work queue and semaphore API results and release all local test
  resources.
• Depends on wqueue: Support custom user work queues. nuttx#19982.

Impact

• New feature: YES, test coverage for custom scheduler and libc user work
  queues.
• User adaptation: NO. The no-argument ostest sequence remains unchanged
  in non-Flat builds; the focused test is selected with ostest wqueue.
• Build process: NO. CONFIG_TESTING_OSTEST_WQUEUE defaults to enabled
  only when its pthread and work queue dependencies are available.
• Hardware: NO production hardware behavior is changed.
• Documentation: NO separate application documentation is required; the
  Kconfig help describes backend selection.
• Security: NO impact.
• Compatibility: Existing Flat ostest work queue coverage is retained and
  expanded.

Testing

Build host:

• Linux 6.8.0-60-generic, x86_64
• GNU Arm Embedded Toolchain 10.3.1 20210824

Target:

• ARM Cortex-M7, STM32H7, PX4 FMUv6C
• Apache nuttx-apps master 48ddaadad1499404addd891190f64cccf50008b2
• Apache NuttX master plus wqueue: Support custom user work queues. nuttx#19982

Before change:

The ostest work queue source is selected only with CONFIG_SCHED_WORKQUEUE.
It has no focused command entry and cannot exercise libc custom user queues
in a Protected user process.

Build output after change:

master_flat:
  FLASH     1470396 B / 1920 KB (74.79%)
  AXI_SRAM    63380 B / 512 KB  (12.09%)

master_protected user:
  uflash      751000 B / 1 MB   (71.62%)
  usram         8192 B / 384 KB (2.08%)

Runtime output summary:

nsh> time "ostest wqueue"

Flat runs:
  30.1490 / 30.1490 / 30.1480 s

Protected libc user runs:
  15.6540 / 15.6540 / 15.6540 s

Every run:
  API validation done
  periodic calls = 3, expect = 3
  replacement total = 2, expect = 2
  sync cancel finished = 1, expect = 1
  parallel callbacks = 2, expect = 2
  multiple queues loop 1/4 through 4/4 done
  self free result = -35, expect = -35
  pending callback = 0, expect = 0
  running callback = 1, expect = 1
  teardown requeue rejected
  teardown done

All three runs for each backend completed without assertion. Current heap
usage did not increase after the first run, and all custom workers exited.

Validation:

checkpatch commit messages: pass
checkpatch complete patch range: pass
full-file nxstyle for wqueue.c and ostest_main.c: pass
cmake-format: pass
git diff --check: pass

PR verification Self-Check

• This PR introduces only one functional change.
• I have updated all required description fields above.
• My PR adheres to the contributing guidelines and coding standard.
• My PR is still work in progress.
• My PR is ready for review and can be safely merged.

[13022591351]

Thanks for the earlier review. I force-pushed the branch to rebase and
organize the follow-up work into two focused commits:

1. testing/ostest: exercise custom work queues

   • Exercises dynamically created queues with one and two workers.
   • Covers invalid arguments, periodic requeue, pending replacement,
     synchronous and parallel cancellation, multiple simultaneous queues,
     self-destruction rejection, and teardown.
   • Replaces unbounded semaphore waits with bounded waits and verifies
     exactly 100 callback executions.

2. testing/ostest: test USRWORK without pthreads

   • Allows the focused WQ test to build with CONFIG_LIBC_USRWORK=y and
     CONFIG_DISABLE_PTHREAD=y.
   • Exercises real work on the predefined USRWORK queue, including
     immediate and delayed work, replacement, cancellation, synchronous
     cancellation, and callback-driven periodic requeue.

The Protected kernel harness is intentionally not included in this apps PR.
The existing non-Flat no-argument ostest flow remains unchanged.

The final two-commit version was validated on an STM32H7 PX4 FMUv6C:

• Flat ostest wqueue: 30.1490 / 30.1480 / 30.1480 s
• Protected libc user WQ: 15.6540 / 15.6540 / 15.6540 s
• Protected USRWORK without pthreads: 0.2290 / 0.2290 / 0.2290 s

All runs completed successfully, and repeated runs showed no cumulative
Kmem or Umem growth. The force-push also normalizes the commit subjects to
the current NuttX style.

[xiaoxiang781216]

why need check __KERNEL__

[xiaoxiang781216]

remove the check

[xiaoxiang781216]

sqaush into the first patch

[13022591351]

Thanks for the review, and sorry for the delayed update.

I force-pushed the branch after addressing all three comments:

• Removed the __KERNEL__ path from the apps test.
• Removed the top-level CONFIG_DISABLE_PTHREAD implementation split. Semaphore waits and basic workers are now shared, while conditional compilation is limited to tests that actually require pthread-backed custom queues.
• Squashed the no-pthread USRWORK coverage into the main test commit.

The branch is now a single commit rebased onto the current nuttx-apps master.
The dependent NuttX work queue API PR, apache/nuttx#19982, has also been merged.

I rebuilt and tested the updated code on PX4 FMUv6C in Flat, Protected, and Protected-with-pthreads-disabled configurations. All work queue tests passed, including repeated runs without cumulative heap growth.