On Mon, May 17, 2021 at 11:47:11AM +0100, Alex Bennée wrote:
Daniel P. Berrangé <berrange(a)redhat.com> writes:
> On Wed, May 05, 2021 at 11:41:46AM +0100, Alex Bennée wrote:
>>
>> Alex Bennée <alex.bennee(a)linaro.org> writes:
>>
>> > Daniel P. Berrangé <berrange(a)redhat.com> writes:
>> >
>> >> On Wed, May 05, 2021 at 10:22:57AM +0100, Alex Bennée wrote:
>> <snip>
>> >>> +TCG introspection features
>> >>> +--------------------------
>> >>> +
>> >>> +TCG trace-events (since 6.1)
>> >>>
+''''''''''''''''''''''''''''
>> >>> +
>> >>> +The ability to add new TCG trace points has bit rotted and as the
>> >>
>> >> When you say this "has bit rotted", just how bad is the
situation ?
>> >>
>> >> Is the TCG tracing still usable at all, or is is fully broken
>> >> already ?
>> >
>> > Well patches 6/7 got it working for generic TCG things. I haven't been
>> > able to get the architecture one working but I suspect that is some sort
>> > of interaction between the per-arch trace header generation that I
>> > haven't quite figured out yet.
>>
>> Ahh it's since 7609ffb919 (trace: fix tcg tracing build breakage) which
>> limited tcg/vcpu events to the root trace-events file.
>
> That commit is from release 2.10.0.
>
> The other commit mentioned in patch 6 (73ff061032) is from 2.12.0.
>
> So no one has been able to use this feature for 3+ years already.
>
> Is it actually worth fixing and then deprecating for 2 releases before
> deleting, as opposed to just deleting the broken code today on basis
> that it can't have any current users ?
Well I can get it up and running with the aforementioned patches and it
seems reasonable to give some notice. I'm happy to defer to Stefan here
though as it's his sub-system.
Lluís Vilanova was the author and probably main user. He mentioned he's
been away from QEMU for a while.
If you want to drop the feature, I think that's fine since it has
already been broken for over 3 years. If someone wants it back then it
can be added via TCG plugins in the future.
Stefan