On 12/14/20 10:56 PM, Philippe Mathieu-Daudé wrote: > Hi Claudio, Eduardo. > > On 12/14/20 8:10 PM, Eduardo Habkost wrote: >> On Sat, Dec 12, 2020 at 04:55:23PM +0100, Claudio Fontana wrote: >>> From: Eduardo Habkost <ehabk...@redhat.com> >>> >>> since tcg_cpu_ops.h is only included in cpu.h, >>> and as a standalone header it is not really useful, >>> as tcg_cpu_ops.h starts requiring cpu.h defines, >>> enums, etc, as well as (later on in the series), >>> additional definitions coming from memattr.h. >>> >>> Therefore rename it to tcg_cpu_ops.h.inc, to warn >>> any potential user that this file is not a standalone >>> header, but rather a partition of cpu.h that is >>> included conditionally if CONFIG_TCG is true. >> >> What's the benefit of moving definitions to a separate file, if >> the new file is not a standalone header? >
the benefit is avoiding a 100 line ifdef CONFIG_TCG, and already separating out what is tcg-specific and what isn't, but if this is a problem we can avoid that, and revisit later on. > Claudio, I haven't been following every respin. If you did that > change just to please me then the circular dependency remarked by > Richard, then if it simplify the series I'm OK if you have to > remove the includes. Richard, From the answer of Philippe and Eduardo I think they are not ok with .h.inc, I think the option of just putting everything in cpu.h was ok with you, should I go with that? Thanks, Claudio > > Eduardo, if you are happy with patches 1-8 (x86 specific), maybe > you can queue them already. The rest is more TCG generic and > will likely go via Richard/Paolo trees IMO. > >> >> If moving the definitions to a separate header is going to >> require too much work, it's completely OK to keep them in cpu.h >> by now, and try to move them later. >> >> I'm worried that the scope of this series is growing too much, >> and discussion/review of additional changes in each new version >> is preventing us from merging the original changes where we >> already had some consensus. >
