Re: [GIT pull] perf/urgent for 5.7-rc2

[Josh Poimboeuf <jpoimboe@redhat.com>]

On Mon, Apr 20, 2020 at 12:21:46PM -0700, Linus Torvalds wrote:
> On Mon, Apr 20, 2020 at 11:17 AM Josh Poimboeuf <jpoimboe@redhat.com> wrote:
> >
> > If you added something bad to a file, and just rebuilt that file, you
> > wouldn't see the objtool warning until later when you build the entire
> > kernel.
> 
> Yes, that's not optimal, but I think I'd personally still prefer that
> behavior. Especially since I seldom build single files, and in fact
> rather seldom build without some "make -j32" or similar: so build
> errors don't happen linearly in the first place.
> 
> > (Of course the same complaint would apply to vmlinux.o
> > validation.)  But the warning shows the .o file, which could be
> > confusing.
> 
> The warning should show the proper loe-level *.o file, so I don't see
> what's confusing about that.
> 
> Yes, the error would happen while trying to link (say)
> kernel/built-in.a, and 'make' would report that creating that archive
> had failed, but 'objtool' itself would report the particular object
> file it was working on that had issues.

My workflow is often

  $ vi kernel/livepatch/core.c
  $ make kernel/livepatch/core.o

That way I don't have to wait to see what I broke (compile or objtool).

And if whatever I changed was in response to an objtool warning for
kernel/livepatch/core.o, and the warning didn't show up after doing the
above, I might stupidly assume I fixed it.

That's what I meant about it being confusing.  Maybe that's just me
though.

> So the errors should be pretty obvious. But like PeterZ, the makefile
> magic escapes me.

Ditto...

-- 
Josh