[PATCH] add build dependency on lxc_protocol.h to remote_daemon

[Michal Prívozník]

On 3/10/22 21:53, Joe Slater wrote:
> remote_daemon.c and others need the generated header lxc_protocol.h,
> but do not have it as a dependency in meson.build.  This means that
> builds will randomly (ok, very occasionally) fail.  Restructure how the
> header is built so that remote_daemon can have it as a dependency.
> 
> Signed-off-by: Joe Slater <joe.slater at windriver.com>
> 
> ---
>  src/remote/meson.build | 48 ++++++++++++++++++++++++------------------
>  1 file changed, 28 insertions(+), 20 deletions(-)
> 
> diff --git a/src/remote/meson.build b/src/remote/meson.build
> index 0a18826..31a30ee 100644
> --- a/src/remote/meson.build
> +++ b/src/remote/meson.build
> @@ -1,27 +1,11 @@
> -remote_driver_sources = [
> -  'remote_driver.c',
> -  'remote_sockets.c',
> -]
> -
> -remote_driver_generated = []
> +remote_xxx_generated = []
>  
>  foreach name : [ 'remote', 'qemu', 'lxc' ]
> -  client_bodies_h = '@0 at _client_bodies.h'.format(name)
>    protocol_c = '@0 at _protocol.c'.format(name)
>    protocol_h = '@0 at _protocol.h'.format(name)
>    protocol_x = '@0 at _protocol.x'.format(name)
>  
> -  remote_driver_generated += custom_target(
> -    client_bodies_h,
> -    input: protocol_x,
> -    output: client_bodies_h,
> -    command: [
> -      gendispatch_prog, '--mode=client', name, name.to_upper(), '@INPUT@',
> -    ],
> -    capture: true,
> -  )
> -
> -  remote_driver_generated += custom_target(
> +  remote_xxx_generated += custom_target(
>      protocol_h,
>      input: protocol_x,
>      output: protocol_h,
> @@ -30,7 +14,7 @@ foreach name : [ 'remote', 'qemu', 'lxc' ]
>      ],
>    )
>  
> -  remote_driver_generated += custom_target(
> +  remote_xxx_generated += custom_target(
>      protocol_c,
>      input: protocol_x,
>      output: protocol_c,
> @@ -42,6 +26,30 @@ foreach name : [ 'remote', 'qemu', 'lxc' ]
>    rpc_probe_files += files(protocol_x)
>  endforeach
>  
> +
> +remote_driver_sources = [
> +  'remote_driver.c',
> +  'remote_sockets.c',
> +]
> +
> +remote_driver_generated =remote_xxx_generated
> +
> +foreach name : [ 'remote', 'qemu', 'lxc' ]
> +  client_bodies_h = '@0 at _client_bodies.h'.format(name)
> +  protocol_x = '@0 at _protocol.x'.format(name)
> +
> +  remote_driver_generated += custom_target(
> +    client_bodies_h,
> +    input: protocol_x,
> +    output: client_bodies_h,
> +    command: [
> +      gendispatch_prog, '--mode=client', name, name.to_upper(), '@INPUT@',
> +    ],
> +    capture: true,
> +  )
> +
> +endforeach
> +
>  remote_daemon_sources = files(
>    'remote_daemon.c',
>    'remote_daemon_config.c',
> @@ -49,7 +57,7 @@ remote_daemon_sources = files(
>    'remote_daemon_stream.c',
>  )
>  
> -remote_daemon_generated = []
> +remote_daemon_generated = remote_xxx_generated

So IIUC, this fix consists of two parts:
1) generating client_bodies_h only AFTER protocol.x for all three
drivers was processed,
2) Initializing remote_daemon_generated to remote_xxx_generated (which
contains protocol.x processing targets).

IMO, it's only the second step that's really needed, isn't it? I'm not
against this patch, I'm just trying to confirm I understood the problem
and the fix.

Michal