Re: [PATCH v3 5/5] dlls/ntdll: allocate a console for initial CUI process when not attached to a Unix console

[Eric Pouech]

Le 20/02/2022 à 13:20, Jacek Caban a écrit :
> Hi Eric,
>
> On 2/17/22 10:10, Eric Pouech wrote:
> > suggested by Jacek Caban
> >
> > Signed-off-by: Eric Pouech
> >
> > dlls/ntdll/unix/env.c |    2 ++
> >   1 file changed, 2 insertions(+)
> >
> > diff --git a/dlls/ntdll/unix/env.c b/dlls/ntdll/unix/env.c
> > index c7e0674e083..a69d0f26e87 100644
> > --- a/dlls/ntdll/unix/env.c
> > +++ b/dlls/ntdll/unix/env.c
> > @@ -1657,6 +1657,8 @@ static void get_initial_console( 
> > RTL_USER_PROCESS_PARAMETERS *params )
> >       wine_server_fd_to_handle( 1, GENERIC_WRITE|SYNCHRONIZE, 
> > OBJ_INHERIT, &params->hStdOutput );
> >       wine_server_fd_to_handle( 2, GENERIC_WRITE|SYNCHRONIZE, 
> > OBJ_INHERIT, &params->hStdError );
> >   +    params->ConsoleHandle = CONSOLE_HANDLE_ALLOC;
> > -
> >
> > /* mark tty handles for kernelbase, see init_console */
> >       if (params->hStdInput && isatty(0))
> >       {
>
> Looking more at this, I'm not sure if it's a good idea, at least in 
> its current form. There are situations where user would not want the 
> console to be created. For example, user may have a Unix script that 
> runs some non-interactive Windows commands with Wine, which is not 
> attached to any console. I guess that some heuristic like if 
> (!params->hStdInput && !params->hStdOutput) should mostly catch such 
> cases, but it's still not perfect. I guess users could still use 
> something like "start /b" to be explicit, but it's not obvious if it's 
> worth it.

agreed (may be you skipped the bottom note in cover mail [PATCH v3 0/5] 
.... https://www.winehq.org/pipermail/wine-devel/2022-February/208328.html)
> BTW, I noticed that start.exe CREATE_NEW_CONSOLE handling will not be 
> right with your patches. "/b" option should probably use 
> DETACHED_PROCESS.

yes. and likely the CREATE_NEW_CONSOLE init is not needed either.

I'll repost

thanks for the review
> Thanks,
>
> Jacek