Testcase code should not directly include "cfe_test_msgids.h" (yet)

[jphickey]

Describe the bug
The module-specific header file is a relatively new addition. But for a user that has overridden the previous all-inclusive cfe_msgids.h file, including the test_msgids.h directly will cause conflict with the overrides.

To Reproduce
Override (only) cfe_msgids.h in a build, and change the msgids in here. Do not override cfe_test_msgids.h.
Attempt to build cfe_testcase -- the cfe_test_msgids.h inclusion will bypass the modified msgids, and get defaults, which will conflict / mismatch.

Expected behavior
Should be backward compatible with an existing override of cfe_msgids.h

System observed on:
Debian

Reporter Info
Joseph Hickey, Vantage Systems, Inc.

[skliper]

Reopening w/ Dan's concurrence, this really should be fixed. Another option is for the modules override search to accept a top level *_msgids.h... then it wouldn't matter since all the generated wrappers would pull in the right file.

Basically it currently searches for:

*_defs/cfe_test_msgids.h
*_defs/default_cfe_test_msgids.h
*_defs/cpu1_cfe_test_msgids.h
*_defs/cpu1/cfe_test_msgids.h
*_defs/cpu1/default_cfe_test_msgids.h
*_defs/default/cfe_test_msgids.h
*_defs/default/cpu1_cfe_test_msgids.h
*_defs/config/cfe_test_msgids.h
*_defs/config/default_cfe_test_msgids.h
*_defs/config/cpu1_cfe_test_msgids.h
*_defs/cpu1/config/cfe_test_msgids.h
*_defs/cpu1/config/default_cfe_test_msgids.h
*_defs/default/config/cfe_test_msgids.h
*_defs/default/config/cpu1_cfe_test_msgids.h*_defs/cfe_test_msgids.h

but the first check should be for a global msgid override (*_defs/cpu1_msgids.h is what we use). Similar pattern for performance ids. Related to

• Avoid generating module-specific msgid files if project has provided an all-inclusive "cfe_msgids.h" file #2405

[jphickey]

@skliper -- Is there a specific problem with your current msgid file? It should still find (and use) a file called cpu1_msgids.h and use that for cfe_msgids.h on cpu1. Can you confirm the content of "cfe_msgids.h" in your build is still as you expected it to be?

And yes - cfe_test_msgids.h (along with all the others) will still be generated. For now, they should be ignored -- nothing includes them, if not using the default msgids.

I will look again into some method of (possibly) avoiding the confusion of generating a file that is not used. But other than that potentially-confusing aspect - is something broken here?

[skliper]

@jphickey - the test code directly references cfe_test_msgids.h, this is what's broken for us (if it was ignored we'd be fine). As this issue states, the test code should not directly include cfe_test_msgids.h if you want to maintain backwards compatibility.

We are currently maintaining a work around by managing an override cfe_test_msgids.h in our *_defs (as just a link to cpu1_msgids.h).

[skliper]

@dzbaker Recommend closing this in favor of #2405 due to the issues this causes per comment on #2404 about breaking builds that don't include cfe_testcase.

#2404 (comment)

[skliper]

Note that I was able to work around my issue by overriding the cfe_test_msgids.h source from target_defs/arch_build_custom.cmake:
set(TEST_CFGFILE_SRC_cfe_test_msgids ${MISSION_DEFS}/cpu1_msgids.h)

The another option would be to create a symbolic link in target_defs from cfe_test_msgids.h to cpu1_msgids.h but that was less desireable.

[dzbaker]

Superseded by #2405.