linker: use wildcards in rel-sections.ld

[dcpleung]

Update rel-sections.ld to use wildcards instead of
spelling out those sections one by one.

Fixes #10493

Signed-off-by: Daniel Leung [email protected]

[codecov-io]

Codecov Report

Merging #10536 into master will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master   #10536   +/-   ##
=======================================
  Coverage   53.37%   53.37%           
=======================================
  Files         210      210           
  Lines       25825    25825           
  Branches     5686     5686           
=======================================
  Hits        13783    13783           
  Misses       9727     9727           
  Partials     2315     2315

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 1a1df58...0678f5c. Read the comment docs.

[aescolar]

shouldnt this be extended to .rel.shell_*.?
Otherwise shell_cmd_ and shell_module_ will be missing.

[dcpleung]

Thanks for letting me know. I was just looking at common-rom.ld.

[pfalcon]

From #10528 (comment)

Good to know. I will update PR #10536 with wildcard ones.

@dcpleung : I wanted to add that to #10493, but apparently forgot: did you consider not doing any .ld file hacking for native_posix at all? Like, make it link with default host toolchain settings. (I don't know if that's actually easier, or we have a custom .ld even for native_posix, which is a portability hazard in general).

[dcpleung]

Thanks to @SebastianBoe for letting me know that ordering works. So updated the patch to use wildcards to match them all. :)

[dcpleung]

@pfalcon We do have a custom linker script for posix at include/arch/posix/linker.ld, but it acts as an addition to the one provided by the host toolchain. Thanks for the note. I have removed it from the posix linker script in the latest revision.

[aescolar]

did you consider not doing any .ld file hacking for native_posix at all?

I have some concerns in the same direction. Given that the posix arch is meant to be used with whatever the user wants/has locally, we have very little control over it. So I'm quite concerned about the amount of issues this may raise over time and the involved maintenance effort for this feature in this arch.
My thinking was more in the line of: "Should this be kconfigurable, so for those who have problems with it, they could easily disable it"

[aescolar]

I have removed it from the posix

Just removing it will cause all the warnings to come back. You would want to ifdef the linker flag in the cmake file too

[dcpleung]

@aescolar I have disabled orphan section warning for native_posix. If this needs to work, we actually have to override the .rel.plt (as we need the ordering to work correctly) and it will probably be problematic in the future.

[aescolar]

@dcpleung Thanks. Question:
In 8ce758a you added quite a few more things to the posix linker script (include/arch/posix/linker.ld). Are those necessary?

[dcpleung]

@aescolar Those are still needed (with GCC 6.4.0 I am using on my machine).

[aescolar]

Those are still needed (with GCC 6.

This is a sidetrack, and not a comment to this PR in particular, but could you describe the problem you had which required them?

[dcpleung]

@aescolar Come to think of it, they are no longer needed as we are not warning about orphan sections for posix anymore... (I need more caffeine... :) )

[dcpleung]

Also reverted include/arch/posix/linker.ld to the version before all the orphan section changes.