Re: [Qemu-trivial] [Qemu-devel] [PULL 21/23] bsd-user: replace fprintf(stderr, ...) with error_report()

[Peter Maydell]

On 2 June 2014 13:08, Michael Tokarev <address@hidden> wrote:
> I applied it and actually tried to compile-check it, on kFreeBSD.
> Compile went successfully, and I was satisfied, until I figured
> that my kFreeBSD test script only builds qemu-system.  So I went
> back and enabled this, and actually found the issue and even
> fixed it locally, by adding the #includes.  While doing this, I
> wondered, why such a basic/common subsystem is not included in
> there to start with, and so isn't used?  Maybe this is something
> which shouldn't be done?

I think it's not inherently wrong to use the error-reporting
infrastructure -- the *-user targets historically were very
much "only use the CPU emulation stuff", but they've gradually
accumulated more use of generic QEMU utility routines, especially
with the advent of QOM. But if we want to do that then we should
probably start with the linux-user code, which gets much more
testing, and only convert bsd-user when we're happy that it
doesn't have any unexpected issues. (Especially since bsd-user
as it stands is badly broken; I'd like to encourage the BSD folks
who are currently working on landing fixes and improvements to
it, so I'd prefer not to give them unnecessary extra rebasing
and merge-conflict resolution work by churning the current
mainline code too much.)

>> In short, I think we need to revert this commit
>> (1fba509527beb).
>
> Yes, that's what I think too.  Should I send a formal patch
> submission, or is `git revert' easy enough?  Even with my
> Signed-off-by: Michael Tokarev <address@hidden>
> if needed?

I'll just do a git-revert, with your acked-by, and push it
to master.

thanks
-- PMM