[PATCH 1/4] ntdll: Init CPU info before calling signal_init_threading.

[Jacek Caban]

Signed-off-by: Jacek Caban 
---
  dlls/ntdll/unix/loader.c       | 2 +-
  dlls/ntdll/unix/system.c       | 4 ++--
  dlls/ntdll/unix/unix_private.h | 2 +-
  3 files changed, 4 insertions(+), 4 deletions(-)

[Alexandre Julliard]

Jacek Caban  writes:
> Signed-off-by: Jacek Caban
>
> dlls/ntdll/unix/loader.c       | 2 +-
>  dlls/ntdll/unix/system.c       | 4 ++--
>  dlls/ntdll/unix/unix_private.h | 2 +-
>  3 files changed, 4 insertions(+), 4 deletions(-)
>
> diff --git a/dlls/ntdll/unix/loader.c b/dlls/ntdll/unix/loader.c
> index 8b1fd6edb7a..45ae435bbc6 100644
> --- a/dlls/ntdll/unix/loader.c
> +++ b/dlls/ntdll/unix/loader.c
> @@ -1643,13 +1643,13 @@ static void start_main_thread(void)
>      NTSTATUS status;
>      TEB *teb = virtual_alloc_first_teb();
> - init_cpu_info( teb );
> signal_init_threading();
> signal_alloc_thread( teb );
> signal_init_thread( teb );
> dbg_init();
> startup_info_size = server_init_process();
> virtual_map_user_shared_data();
> - init_cpu_info();
> init_files();
> init_startup_info();
> NtCreateKeyedEvent( &keyed_event, GENERIC_READ | GENERIC_WRITE, NULL, 0 );

That works of course, but I'd find it cleaner to keep the thread init
before anything else.  There's no reason that the syscall dispatcher
needs to be setup at that point, we can have a signal_init_syscalls()
function called later on.

[Jacek Caban]

On 19.02.2021 18:26, Alexandre Julliard wrote:
> Jacek Caban  writes:
> > Signed-off-by: Jacek Caban
> >
> > dlls/ntdll/unix/loader.c       | 2 +-
> >   dlls/ntdll/unix/system.c       | 4 ++--
> >   dlls/ntdll/unix/unix_private.h | 2 +-
> >   3 files changed, 4 insertions(+), 4 deletions(-)
> >
> > diff --git a/dlls/ntdll/unix/loader.c b/dlls/ntdll/unix/loader.c
> > index 8b1fd6edb7a..45ae435bbc6 100644
> > --- a/dlls/ntdll/unix/loader.c
> > +++ b/dlls/ntdll/unix/loader.c
> > @@ -1643,13 +1643,13 @@ static void start_main_thread(void)
> >       NTSTATUS status;
> >       TEB *teb = virtual_alloc_first_teb();
> > - init_cpu_info( teb );
> >  signal_init_threading();
> >  signal_alloc_thread( teb );
> >  signal_init_thread( teb );
> >  dbg_init();
> >  startup_info_size = server_init_process();
> >  virtual_map_user_shared_data();
> > - init_cpu_info();
> >  init_files();
> >  init_startup_info();
> >  NtCreateKeyedEvent( &keyed_event, GENERIC_READ | GENERIC_WRITE, NULL, 0 );
>
> That works of course, but I'd find it cleaner to keep the thread init
> before anything else.  There's no reason that the syscall dispatcher
> needs to be setup at that point, we can have a signal_init_syscalls()
> function called later on.

That looks indeed nicer, I sent a new version. Thanks for reviews.

Jacek