Page MenuHomeFreeBSD

depend-cleanup.sh: Simplify handling for stale syscall.S
ClosedPublic

Authored by jhb on Jul 24 2024, 4:33 PM.
Tags
None
Referenced Files
Unknown Object (File)
Oct 2 2024, 7:19 AM
Unknown Object (File)
Oct 1 2024, 7:41 PM
Unknown Object (File)
Sep 30 2024, 12:31 AM
Unknown Object (File)
Sep 29 2024, 10:05 PM
Unknown Object (File)
Sep 29 2024, 8:12 AM
Unknown Object (File)
Sep 29 2024, 8:11 AM
Unknown Object (File)
Sep 18 2024, 7:17 AM
Unknown Object (File)
Sep 16 2024, 5:58 AM
Subscribers

Details

Summary

Remove the outer grep and depend on the greps in clean_dep instead.

Obtained from: CheriBSD

Diff Detail

Repository
rG FreeBSD src repository
Lint
Lint Skipped
Unit
Tests Skipped
Build Status
Buildable 58805
Build 55692: arc lint + arc unit

Event Timeline

jhb requested review of this revision.Jul 24 2024, 4:33 PM

This will remove a valid depends file for the lib32 builds on amd64 on each build. The issue I ran into was that when upgrading from a pre-libsys tree to a libsys tree on a arch with other compat arches, the other compat arches weren't cleaned and failed.

I was worried this would always force a rebuild of syscall.o.
But reading clean_dep I've convinced myself that this is a good change.
And reason to not also do it on i386? Why do we still need the MACHINE check?

This revision is now accepted and ready to land.Jul 24 2024, 5:02 PM
In D46103#1050896, @imp wrote:

I was worried this would always force a rebuild of syscall.o.
But reading clean_dep I've convinced myself that this is a good change.
And reason to not also do it on i386? Why do we still need the MACHINE check?

i386 still has syscall.S in its libc. Hence why lib32 on amd64 will also still have syscall.S.

We could decide that i386 doesn't need to be weird and drop the assembly. I didn't do it with the others because it is a modest optimization to call the intended syscall directly from userspace (vs other architectures where it was just a pointless extra file.)