Bug fix PR nearly two months without a review. Posting a useless comment to keep your bot from marking it stale.

@aclamk @ifed01 Does this look reasonable to you?

It's been another month.

@MaxKellermann hi Max, thank you for this fix. but i think the reason why we linked os against blk was because of the bluestore backend, which in turn uses BlockDevice to access the underlying block devices. so instead of generating an interface library for blk, i think a better fix would be to put something like:

if(WITH_BLUESTORE)
  target_link_libraries(os
    blk)
endif()

where we don't link os against blk unconditionally. this change makes the dependency more obvious. what do you think?

@tchaikov, that would be another way to fix this build failure; but I decided against it because it would require fixes in 3 places (src/crimson/os/alienstore/CMakeLists.txt, src/librbd/CMakeLists.txt, src/os/CMakeLists.txt), adding 3 more cmake conditionals; while my approach adds zero conditionals, as it only adds an else to an existing conditional. It's the simplest possible fix for this bug.
Under these circumstances, would you want me to change this PR to implement the other approach you suggested?

thanks, Max. i will take a closer look at it and get back to you in this week.

@tchaikov, that would be another way to fix this build failure; but I decided against it because it would require fixes in 3 places (src/crimson/os/alienstore/CMakeLists.txt, src/librbd/CMakeLists.txt, src/os/CMakeLists.txt), adding 3 more cmake conditionals; while my approach adds zero conditionals, as it only adds an else to an existing conditional. It's the simplest possible fix for this bug. Under these circumstances, would you want me to change this PR to implement the other approach you suggested?

a dummy library is indeed simpler. thanks!

Can one of the admins verify this patch?

jenkins test api

jenkins test make check arm64