Re: [PATCH v7 02/23] cpu: Restrict target cpu_do_transaction_failed() handlers to sysemu - Philippe Mathieu-Daudé

From: "Philippe Mathieu-Daudé" <f4bug@amsat.org>
To: Richard Henderson <richard.henderson@linaro.org>, qemu-devel@nongnu.org
Cc: qemu-riscv@nongnu.org, Laurent Vivier <laurent@vivier.eu>,
	qemu-s390x@nongnu.org, qemu-arm@nongnu.org, qemu-ppc@nongnu.org,
	Paolo Bonzini <pbonzini@redhat.com>
Subject: Re: [PATCH v7 02/23] cpu: Restrict target cpu_do_transaction_failed() handlers to sysemu
Date: Wed, 26 May 2021 19:22:21 +0200	[thread overview]
Message-ID: <5a1389f6-ced6-3b09-ba9b-749e2db2de4b@amsat.org> (raw)
In-Reply-To: <ae262a9c-1ffa-b680-4aa6-d7256fa6947a@linaro.org>

On 5/26/21 4:12 AM, Richard Henderson wrote:
> On 5/17/21 3:51 AM, Philippe Mathieu-Daudé wrote:
>> In commit cbc183d2d9f ("cpu: move cc->transaction_failed to tcg_ops")
>> we restricted the do_transaction_failed() handler to the sysemu part
>> of TCGCPUOps, but forgot to restrict the target specific declarations.
>>
>> Signed-off-by: Philippe Mathieu-Daudé<f4bug@amsat.org>
>> ---
>>   target/arm/internals.h |  2 ++
>>   target/m68k/cpu.h      |  2 ++
>>   target/riscv/cpu.h     | 10 +++++-----
>>   target/xtensa/cpu.h    |  8 ++++----
>>   4 files changed, 13 insertions(+), 9 deletions(-)
> 
> What do the extra ifdefs buy us? Surely the fact that the symbol is not
> present in the user-only, and would produce link errors if used, is
> sufficient?

Last month on another series you told me we should declare prototypes
that can't be used :) However I agree we want *less* #ifdef'ry, not
more.

The ARM declarations is fixed in Claudio's accel rework.

For the m68k/riscv/xtensa targets I'll see if I can move it elsewhere
in the same file (if there is already such #ifdef'ry), it is worthwhile
to move the declaration in another (sysemu) header, or simply do
nothing.

Thanks,

Phil.