bug#73469: 29.4; mouse passing on terminal window generates events

[Jared Finder via "Bug reports for GNU Emacs, the Swiss army knife of text editors" <bug-gnu-emacs@gnu.org>]

On 2024-09-29 13:07, Francesco Potortì wrote:
> 
> This is under Screen
> ________________________________________________________________
> Sun Sep 29 21:38:26 CEST 2024
> 
> ~$ echo -e "\e[?1000h"
> 
> ~$ echo -e "\e[?1003h"
> 
> ~$ echo -e "\e[?1006h"
> 
> ~$ 
> #35;40;5M35;40;5M35;40;5M35;40;5M35;40;5M35;40;5M35;40;5M35;40;5M35;40;5M35;40;6M35;40;6M35;40;7M35;40;7M35;40;7M35;40;7M35;40;7M35;40;8M35;40;8M35;40;8M35;40;8M35;40;8M35;40;8M35;40;8M35;40;9M35;40;9M
> ~$ # above, I just moved the mouse pointer

This is odd and indicates of a bug outside of Emacs.

A mouse move with no buttons down should be sending lowercase m, not 
uppercase M. This is aligned with what the Emacs lossage reported, 
though we can't see the initial ESC [ M characters as the terminal tries 
to interpret that.

> ~$ # now, I will click left, then right, then center
> ~$ #0;41;9M0;41;9m35;41;9M35;41;9M1;41;9M1;41;9m
> ~$ # right did not work, as it is taken by mate-terminal

This may be ok; it depends on if it is prefixed with ESC [ M (bad!) or 
ESC [ < (good!).

> ~$ # now, I rotate the wheel
> ~$ # 
> 64;41;9M64;41;9M64;41;9M65;41;9M65;41;9M65;41;9M65;41;9M64;41;9M64;41;9M65;41;9M65;41;9M

This is also odd.

The wheel is supposed to look mouse buttons 4 and 5 which is accurately 
represented here. However, the terminal is only sending button down 
(suffixed with M) and no button up (suffixed with m).

> ~$ # now, disabling mode 1015
> ~$  echo -e "\e[?1015l"

All the events here looked the same.

> Now the same on the same terminal, but outside of Screen

Again, all the events here looked the same, though like before I 
couldn't see if the prefix was ESC [ M (bad!) or ESC [ < (good!).

Thank you so much for all this testing. I'm fairly confident this is not 
an issue in Emacs and probably not in screen as well. That means we can 
rely on view-lossage for accurate character reporting to diagnose what's 
going on, which is nice because we can see the escape sequence prefix.


For next steps I am considering adding a workaround and want to verify 
the behavior is easy to implement and maintain. Unfortunately since I 
can't reproduce your issue locally, I do need more info from you if you 
have time. I want to see those escape sequence prefixes for both today 
and the state prior 0695c9e8599b5036a80361571e7cb0ea9fdead99 from 2020 
which added mouse movement tracking.

Can you please report the lossage for the same mouse move, mouse button, 
and mouse wheel test within Emacs and report the lossage? You already 
did mouse move earlier, so I'm just looking for mouse button click and 
mouse wheel scroll.

After that, can you please try altering xterm-mouse--tracking-sequence, 
replacing 1003 with 1002 before enabling xterm-mouse-mode. (This was the 
state prior to mouse movement tracking being supported.) Please do this 
in a fresh Emacs session before enabling xterm-mouse-mode the first 
time. Then report the same mouse move (I'm expecting no characters 
here), mouse click, and mouse wheel lossage.

Thank you!

   -- MJF