Re: [PATCH v4 6/7] dump: Build once by adding stubs for non-x86 targets

[Richard Henderson]

On 2/23/23 13:17, Philippe Mathieu-Daudé wrote:
> Extract non-x86 stubs to win_dump-stub.c. We can now
> build dump.o once for system emulation. Update meson.
>
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   dump/dump.c          | 14 --------------
>   dump/meson.build     |  6 ++++--
>   dump/win_dump-stub.c | 23 +++++++++++++++++++++++
>   3 files changed, 27 insertions(+), 16 deletions(-)
>   create mode 100644 dump/win_dump-stub.c
>
> diff --git a/dump/dump.c b/dump/dump.c
> index b33a613d45..7cde3e326e 100644
> --- a/dump/dump.c
> +++ b/dump/dump.c
> @@ -32,20 +32,6 @@
>   #include "migration/blocker.h"
>   #include "win_dump.h"
>   
> -#ifndef TARGET_X86_64
> -bool win_dump_available(Error **errp)
> -{
> -    error_setg(errp, "Windows dump is only available for x86-64");
> -
> -    return false;
> -}
> -
> -void create_win_dump(DumpState *s, Error **errp)
> -{
> -    win_dump_available(errp);
> -}
> -#endif
> -
>   #include <zlib.h>
>   #ifdef CONFIG_LZO
>   #include <lzo/lzo1x.h>
> diff --git a/dump/meson.build b/dump/meson.build
> index 2eff29c3ea..6ae07e6fed 100644
> --- a/dump/meson.build
> +++ b/dump/meson.build
> @@ -1,4 +1,6 @@
>   softmmu_ss.add(files('dump-hmp-cmds.c'))
>   
> -specific_ss.add(when: 'CONFIG_SOFTMMU', if_true: [files('dump.c'), snappy, lzo])
> -specific_ss.add(when: ['CONFIG_SOFTMMU', 'TARGET_X86_64'], if_true: 
> files('win_dump.c'))
> +softmmu_ss.add(when: 'CONFIG_SOFTMMU', if_true: [files('dump.c'), snappy, lzo])
> +specific_ss.add(when: ['CONFIG_SOFTMMU', 'TARGET_X86_64'],
> +                if_true: files('win_dump.c'),
> +                if_false: files('win_dump-stub.c'))

Doesn't this add win_dump-stub.c when !(SOFTMMU && X86_64), i.e. !SOFTMMU || 
!X86_64?

I trying to imagine how well this will scale with ARM64, for the ongoing Windows on ARM 
project.  Would it just be easier have the stubs in win_dump.c, using ifdefs?


r~