This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH 0/7] Regcache: Split out target regcache functionality
- From: Yao Qi <qiyaoltc at gmail dot com>
- To: Alan Hayward <Alan dot Hayward at arm dot com>
- Cc: "gdb-patches\@sourceware.org" <gdb-patches at sourceware dot org>, nd <nd at arm dot com>
- Date: Wed, 23 Aug 2017 11:02:57 +0100
- Subject: Re: [PATCH 0/7] Regcache: Split out target regcache functionality
- Authentication-results: sourceware.org; auth=none
- References: <D453EE50-9381-49C4-8B02-3EA52138D387@arm.com>
Alan Hayward <Alan.Hayward@arm.com> writes:
> A target_regcache is a regcache connected to a target. Reads and
> writes of register values are passed through to the target.
> A target_regcache cannot be readonly (because this doesn't make
> sense).
At this stage, can we don't assume the readonly-ness of target_regcache?
https://sourceware.org/ml/gdb-patches/2017-07/msg00252.html
>
> Meanwhile, a regcache (sometime referred to as a detached regcache)
> is not connected to a target. By default it is readonly, but does
> not have to be. In addition to the raw registers a regcache also
> caches cooked register values. Duplicating a target_regcache will
> always result in a detached regcache.
>
> Both regcache and target_regcache support the full set of raw_read,
> raw_write, raw_collect, raw_supply, cooked_read, cooked_write
> functions and all their variations. A user of a regache does not
> need to treat the two types any differently - your regcache just
> does the correct thing. For example, on a target_regcache,
> raw_write will write to both the cache and the target, but on
> a recache it will only write to the cache.
>
> With this set of patches, the regcache for the current thread (as
> given by get_current_regcache()), is now a target_regcache, that
> will perform exactly the same functionality as the regcache in
> the exisiting head.
>
> An earlier plan for this set of patches was that the detached
> regcache would not support raw_write, raw_collect, cooked_read,
> cooked_write. The problem is many of the target hooks then need
My suggestion was that detached regcache doesn't have
{raw,cooked}_{read,write}_ methods, only has collect and supply methods.
and attached one has {raw,cooked}_{read,write}_ methods.
https://sourceware.org/ml/gdb-patches/2017-07/msg00127.html
> updating to only accept target_regcache. Many of the implementations
> of the target hooks then call back into the regcache, causing those
> functions in turn to be dependant on the new classes, exploding the
> size of the patch. It then become very fiddly/confusing to maintain
> which type of regcache is required at any point.
target hooks should call regcache supply and collect methods, IMO.
--
Yao (齐尧)