* Re: master 24d01af3dd: Cancel DND correctly during display disconnect on a different display
[not found] ` <20220429012907.00486C01683@vcs2.savannah.gnu.org>
@ 2022-05-26 14:34 ` Ken Brown
2022-05-26 16:34 ` Ken Brown
0 siblings, 1 reply; 3+ messages in thread
From: Ken Brown @ 2022-05-26 14:34 UTC (permalink / raw)
To: emacs-devel, Po Lu
On 4/28/2022 9:29 PM, Po Lu via Mailing list for Emacs changes wrote:
> branch: master
> commit 24d01af3dde22ccd03b614d981129c6eacb0d56e
> Author: Po Lu <luangruo@yahoo.com>
> Commit: Po Lu <luangruo@yahoo.com>
>
> Cancel DND correctly during display disconnect on a different display
>
> * src/xterm.c (x_dnd_io_error_handler): New function.
> (x_connection_closed): Send the messages to cancel DND if a
> different display was disconnected.
> ---
> src/xterm.c | 78 ++++++++++++++++++++++++++++++++++++++++++++++++++++++-------
> 1 file changed, 70 insertions(+), 8 deletions(-)
>
> diff --git a/src/xterm.c b/src/xterm.c
> index c4e0305a6e..35f6bc7fe8 100644
> --- a/src/xterm.c
> +++ b/src/xterm.c
> @@ -1005,6 +1005,10 @@ static int x_dnd_movement_x, x_dnd_movement_y;
> static unsigned int x_dnd_keyboard_state;
> #endif
>
> +/* jmp_buf that gets us out of the IO error handler if an error occurs
> + terminating DND as part of the display disconnect handler. */
> +static jmp_buf x_dnd_disconnect_handler;
> +
> struct x_client_list_window
> {
> Window window;
> @@ -2571,8 +2575,17 @@ x_dnd_compute_toplevels (struct x_display_info *dpyinfo)
> return 0;
> }
>
> -#define X_DND_SUPPORTED_VERSION 5
> +static _Noreturn int
> +x_dnd_io_error_handler (Display *display)
> +{
> +#ifdef USE_GTK
> + emacs_abort ();
> +#else
> + longjmp (x_dnd_disconnect_handler, 1);
> +#endif
> +}
>
> +#define X_DND_SUPPORTED_VERSION 5
>
> static int x_dnd_get_window_proto (struct x_display_info *, Window);
> static Window x_dnd_get_window_proxy (struct x_display_info *, Window);
> @@ -20027,6 +20040,9 @@ x_connection_closed (Display *dpy, const char *error_message, bool ioerror)
> struct x_display_info *dpyinfo = x_display_info_for_display (dpy);
> Lisp_Object frame, tail;
> specpdl_ref idx = SPECPDL_INDEX ();
> + void *io_error_handler;
> + xm_drop_start_message dmsg;
> + struct frame *f;
>
> error_msg = alloca (strlen (error_message) + 1);
> strcpy (error_msg, error_message);
> @@ -20034,16 +20050,61 @@ x_connection_closed (Display *dpy, const char *error_message, bool ioerror)
> /* Inhibit redisplay while frames are being deleted. */
> specbind (Qinhibit_redisplay, Qt);
>
> - /* If drag-and-drop is in progress and the DND frame's display is
> - DPY, cancel drag-and-drop. Don't reset event masks or try to
> - send responses to other programs because the display is going
> + /* If drag-and-drop is in progress, cancel drag-and-drop. If DND
> + frame's display is DPY, don't reset event masks or try to send
> + responses to other programs because the display is going
> away. */
>
> - if ((x_dnd_in_progress || x_dnd_waiting_for_finish)
> - && dpy == (x_dnd_waiting_for_finish
> - ? x_dnd_finish_display
> - : FRAME_X_DISPLAY (x_dnd_frame)))
> + if (x_dnd_in_progress || x_dnd_waiting_for_finish)
> {
> + /* Handle display disconnect errors here because this function
> + is not reentrant at this particular spot. */
> + io_error_handler = XSetIOErrorHandler (x_dnd_io_error_handler);
> +
> + if (!sigsetjmp (x_dnd_disconnect_handler, 1)
> + && x_dnd_in_progress
> + && dpy != (x_dnd_waiting_for_finish
> + ? x_dnd_finish_display
> + : FRAME_X_DISPLAY (x_dnd_frame)))
> + {
> + /* Clean up drag and drop if the drag frame's display isn't
> + the one being disconnected. */
> + f = x_dnd_frame;
> +
> + if (x_dnd_last_seen_window != None
> + && x_dnd_last_protocol_version != -1)
> + x_dnd_send_leave (x_dnd_frame,
> + x_dnd_last_seen_window);
> + else if (x_dnd_last_seen_window != None
> + && !XM_DRAG_STYLE_IS_DROP_ONLY (x_dnd_last_motif_style)
> + && x_dnd_last_motif_style != XM_DRAG_STYLE_NONE
> + && x_dnd_motif_setup_p)
> + {
> + dmsg.reason = XM_DRAG_REASON (XM_DRAG_ORIGINATOR_INITIATOR,
> + XM_DRAG_REASON_DROP_START);
> + dmsg.byte_order = XM_BYTE_ORDER_CUR_FIRST;
> + dmsg.timestamp = FRAME_DISPLAY_INFO (f)->last_user_time;
> + dmsg.side_effects
> + = XM_DRAG_SIDE_EFFECT (xm_side_effect_from_action (FRAME_DISPLAY_INFO (f),
> + x_dnd_wanted_action),
> + XM_DROP_SITE_VALID,
> + xm_side_effect_from_action (FRAME_DISPLAY_INFO (f),
> + x_dnd_wanted_action),
> + XM_DROP_ACTION_DROP_CANCEL);
> + dmsg.x = 0;
> + dmsg.y = 0;
> + dmsg.index_atom = FRAME_DISPLAY_INFO (f)->Xatom_XdndSelection;
> + dmsg.source_window = FRAME_X_WINDOW (f);
> +
> + x_dnd_send_xm_leave_for_drop (FRAME_DISPLAY_INFO (f), f,
> + x_dnd_last_seen_window, 0);
> + xm_send_drop_message (FRAME_DISPLAY_INFO (f), FRAME_X_WINDOW (f),
> + x_dnd_last_seen_window, &dmsg);
> + }
> + }
> +
> + XSetIOErrorHandler (io_error_handler);
> +
> x_dnd_last_seen_window = None;
> x_dnd_last_seen_toplevel = None;
> x_dnd_in_progress = false;
> @@ -20234,6 +20295,7 @@ x_io_error_quitter (Display *display)
> DisplayString (display));
> x_connection_closed (display, buf, true);
> }
> +
> \f
> /* Changing the font of the frame. */
>
>
After this commit I'm getting the following warning on 64-bit Cygwin:
../../master/src/xterm.c:20064:12: warning: initialization of ‘long int (*)[34]’
from incompatible pointer type ‘long int (*)[32]’ [-Wincompatible-pointer-types]
20064 | if (!sigsetjmp (x_dnd_disconnect_handler, 1)
| ^~~~~~~~~
A second such warning appeared after commit cc297af:
../../master/src/xterm.c: In function ‘xm_get_drag_window’:
../../master/src/xterm.c:1447:11: warning: initialization of ‘long int (*)[34]’
from incompatible pointer type ‘long int (*)[32]’ [-Wincompatible-pointer-types]
1447 | if (sigsetjmp (x_dnd_disconnect_handler, 1))
| ^~~~~~~~~
With current git HEAD, the line numbers of the sigsetjmp calls are 1664 and 21315.
Ken
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: master 24d01af3dd: Cancel DND correctly during display disconnect on a different display
2022-05-26 14:34 ` master 24d01af3dd: Cancel DND correctly during display disconnect on a different display Ken Brown
@ 2022-05-26 16:34 ` Ken Brown
2022-05-27 0:10 ` Po Lu
0 siblings, 1 reply; 3+ messages in thread
From: Ken Brown @ 2022-05-26 16:34 UTC (permalink / raw)
To: emacs-devel, Po Lu
On 5/26/2022 10:34 AM, Ken Brown wrote:
>
>
> On 4/28/2022 9:29 PM, Po Lu via Mailing list for Emacs changes wrote:
>> branch: master
>> commit 24d01af3dde22ccd03b614d981129c6eacb0d56e
>> Author: Po Lu <luangruo@yahoo.com>
>> Commit: Po Lu <luangruo@yahoo.com>
>>
>> Cancel DND correctly during display disconnect on a different display
>> * src/xterm.c (x_dnd_io_error_handler): New function.
>> (x_connection_closed): Send the messages to cancel DND if a
>> different display was disconnected.
>> ---
>> src/xterm.c | 78 ++++++++++++++++++++++++++++++++++++++++++++++++++++++-------
>> 1 file changed, 70 insertions(+), 8 deletions(-)
>>
>> diff --git a/src/xterm.c b/src/xterm.c
>> index c4e0305a6e..35f6bc7fe8 100644
>> --- a/src/xterm.c
>> +++ b/src/xterm.c
>> @@ -1005,6 +1005,10 @@ static int x_dnd_movement_x, x_dnd_movement_y;
>> static unsigned int x_dnd_keyboard_state;
>> #endif
>> +/* jmp_buf that gets us out of the IO error handler if an error occurs
>> + terminating DND as part of the display disconnect handler. */
>> +static jmp_buf x_dnd_disconnect_handler;
>> +
>> struct x_client_list_window
>> {
>> Window window;
>> @@ -2571,8 +2575,17 @@ x_dnd_compute_toplevels (struct x_display_info *dpyinfo)
>> return 0;
>> }
>> -#define X_DND_SUPPORTED_VERSION 5
>> +static _Noreturn int
>> +x_dnd_io_error_handler (Display *display)
>> +{
>> +#ifdef USE_GTK
>> + emacs_abort ();
>> +#else
>> + longjmp (x_dnd_disconnect_handler, 1);
>> +#endif
>> +}
>> +#define X_DND_SUPPORTED_VERSION 5
>> static int x_dnd_get_window_proto (struct x_display_info *, Window);
>> static Window x_dnd_get_window_proxy (struct x_display_info *, Window);
>> @@ -20027,6 +20040,9 @@ x_connection_closed (Display *dpy, const char
>> *error_message, bool ioerror)
>> struct x_display_info *dpyinfo = x_display_info_for_display (dpy);
>> Lisp_Object frame, tail;
>> specpdl_ref idx = SPECPDL_INDEX ();
>> + void *io_error_handler;
>> + xm_drop_start_message dmsg;
>> + struct frame *f;
>> error_msg = alloca (strlen (error_message) + 1);
>> strcpy (error_msg, error_message);
>> @@ -20034,16 +20050,61 @@ x_connection_closed (Display *dpy, const char
>> *error_message, bool ioerror)
>> /* Inhibit redisplay while frames are being deleted. */
>> specbind (Qinhibit_redisplay, Qt);
>> - /* If drag-and-drop is in progress and the DND frame's display is
>> - DPY, cancel drag-and-drop. Don't reset event masks or try to
>> - send responses to other programs because the display is going
>> + /* If drag-and-drop is in progress, cancel drag-and-drop. If DND
>> + frame's display is DPY, don't reset event masks or try to send
>> + responses to other programs because the display is going
>> away. */
>> - if ((x_dnd_in_progress || x_dnd_waiting_for_finish)
>> - && dpy == (x_dnd_waiting_for_finish
>> - ? x_dnd_finish_display
>> - : FRAME_X_DISPLAY (x_dnd_frame)))
>> + if (x_dnd_in_progress || x_dnd_waiting_for_finish)
>> {
>> + /* Handle display disconnect errors here because this function
>> + is not reentrant at this particular spot. */
>> + io_error_handler = XSetIOErrorHandler (x_dnd_io_error_handler);
>> +
>> + if (!sigsetjmp (x_dnd_disconnect_handler, 1)
>> + && x_dnd_in_progress
>> + && dpy != (x_dnd_waiting_for_finish
>> + ? x_dnd_finish_display
>> + : FRAME_X_DISPLAY (x_dnd_frame)))
>> + {
>> + /* Clean up drag and drop if the drag frame's display isn't
>> + the one being disconnected. */
>> + f = x_dnd_frame;
>> +
>> + if (x_dnd_last_seen_window != None
>> + && x_dnd_last_protocol_version != -1)
>> + x_dnd_send_leave (x_dnd_frame,
>> + x_dnd_last_seen_window);
>> + else if (x_dnd_last_seen_window != None
>> + && !XM_DRAG_STYLE_IS_DROP_ONLY (x_dnd_last_motif_style)
>> + && x_dnd_last_motif_style != XM_DRAG_STYLE_NONE
>> + && x_dnd_motif_setup_p)
>> + {
>> + dmsg.reason = XM_DRAG_REASON (XM_DRAG_ORIGINATOR_INITIATOR,
>> + XM_DRAG_REASON_DROP_START);
>> + dmsg.byte_order = XM_BYTE_ORDER_CUR_FIRST;
>> + dmsg.timestamp = FRAME_DISPLAY_INFO (f)->last_user_time;
>> + dmsg.side_effects
>> + = XM_DRAG_SIDE_EFFECT (xm_side_effect_from_action (FRAME_DISPLAY_INFO
>> (f),
>> + x_dnd_wanted_action),
>> + XM_DROP_SITE_VALID,
>> + xm_side_effect_from_action (FRAME_DISPLAY_INFO (f),
>> + x_dnd_wanted_action),
>> + XM_DROP_ACTION_DROP_CANCEL);
>> + dmsg.x = 0;
>> + dmsg.y = 0;
>> + dmsg.index_atom = FRAME_DISPLAY_INFO (f)->Xatom_XdndSelection;
>> + dmsg.source_window = FRAME_X_WINDOW (f);
>> +
>> + x_dnd_send_xm_leave_for_drop (FRAME_DISPLAY_INFO (f), f,
>> + x_dnd_last_seen_window, 0);
>> + xm_send_drop_message (FRAME_DISPLAY_INFO (f), FRAME_X_WINDOW (f),
>> + x_dnd_last_seen_window, &dmsg);
>> + }
>> + }
>> +
>> + XSetIOErrorHandler (io_error_handler);
>> +
>> x_dnd_last_seen_window = None;
>> x_dnd_last_seen_toplevel = None;
>> x_dnd_in_progress = false;
>> @@ -20234,6 +20295,7 @@ x_io_error_quitter (Display *display)
>> DisplayString (display));
>> x_connection_closed (display, buf, true);
>> }
>> +
>> \f
>> /* Changing the font of the frame. */
>>
>
> After this commit I'm getting the following warning on 64-bit Cygwin:
>
> ../../master/src/xterm.c:20064:12: warning: initialization of ‘long int (*)[34]’
> from incompatible pointer type ‘long int (*)[32]’ [-Wincompatible-pointer-types]
> 20064 | if (!sigsetjmp (x_dnd_disconnect_handler, 1)
> | ^~~~~~~~~
>
> A second such warning appeared after commit cc297af:
>
> ../../master/src/xterm.c: In function ‘xm_get_drag_window’:
> ../../master/src/xterm.c:1447:11: warning: initialization of ‘long int (*)[34]’
> from incompatible pointer type ‘long int (*)[32]’ [-Wincompatible-pointer-types]
> 1447 | if (sigsetjmp (x_dnd_disconnect_handler, 1))
> | ^~~~~~~~~
>
>
> With current git HEAD, the line numbers of the sigsetjmp calls are 1664 and 21315.
The problem seems to be that x_dnd_disconnect_handler is of type jmp_buf instead
of sigjmp_buf. But we can't simply change the type, because
x_dnd_disconnect_handler is also used in a call to longjmp, which expects an
argument of type jmp_buf.
Ken
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: master 24d01af3dd: Cancel DND correctly during display disconnect on a different display
2022-05-26 16:34 ` Ken Brown
@ 2022-05-27 0:10 ` Po Lu
0 siblings, 0 replies; 3+ messages in thread
From: Po Lu @ 2022-05-27 0:10 UTC (permalink / raw)
To: Ken Brown; +Cc: emacs-devel
Ken Brown <kbrown@cornell.edu> writes:
> The problem seems to be that x_dnd_disconnect_handler is of type
> jmp_buf instead of sigjmp_buf. But we can't simply change the type,
> because x_dnd_disconnect_handler is also used in a call to longjmp,
> which expects an argument of type jmp_buf.
That should be fixed shortly. Sorry for the embarassing mixup.
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2022-05-27 0:10 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
[not found] <165119574603.20331.4036366960372072878@vcs2.savannah.gnu.org>
[not found] ` <20220429012907.00486C01683@vcs2.savannah.gnu.org>
2022-05-26 14:34 ` master 24d01af3dd: Cancel DND correctly during display disconnect on a different display Ken Brown
2022-05-26 16:34 ` Ken Brown
2022-05-27 0:10 ` Po Lu
Code repositories for project(s) associated with this public inbox
https://git.savannah.gnu.org/cgit/emacs.git
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).