06:41 <krobelus> ok so the workaround of clearing leaked entries in the progressive-enhancement flag stack doesn't really help much because it also tramples on interactive commands like "printf '\x1b[>5u'"

06:43 <krobelus> so the advantage of using the stack in the first place kind of goes away. AFAICT that advantage is to allow two sets of processes that access the terminal in an interleaving way to toggle between two sets of flags.

06:45 <krobelus> wouldn't it be simpler if every process set its requisite flags whenever it gains control of the terminal? Then we won't have inconsistent stack state after crashes in some cases.

06:48 <krobelus> to make an interactive "printf '\x1b[=5u'" do the right thing, the shell can query progressive enhancement flags after every command. Unless the command failed perhaps; in that case reset to 0 (for crashes / connection drops)

06:50 <dnkl> yeah, might be easier if applications, instead of pushing on startup, and popping on exit, just reset the flags to 0 on exit, requiring e.g the shell to re-enable them when it regains control

06:52 <dnkl> krobelus: unrelated question: if I'm not mistaken, fish uses OSC-0 to set the window title. Why doesn't it use OSC-2 instead?

06:52 <dnkl> osc-0 also sets the icon, which gets weird when the string is e.g CWD...

07:09 <krobelus> dnkl: apparently because they want to set a "tab title" in addition to the window title: https://github.com/fish-shell/fish-shell/issues/47

07:10 <dnkl> hmm, alright... so iTerm stole an escape that's been defined by xterm since forever...

07:10 <dnkl> great

07:11 <dnkl> thanks for digging though!

07:11 <krobelus> where does the icon title usually show up?

07:17 <dnkl> if I'm reading xterm correctly, it sets the _NET_WM_ICON_NAME property on the X window. Window managers should use it to lookup which icon to display

07:17 <dnkl> assuming I read all the specs correctly

07:18 <dnkl> i.e it does what you'd expect - change the icon associated with the running instance

07:18 <dnkl> I'm planning on implementing it with the new xdg-toplevel-icon protocol

07:20 <krobelus> we can set the icon title only for terminals that need it (iterm and possibly tmux)

07:23 <dnkl> if I remember correctly, Konsole also supports setting the tab title, but they use OSC-30...

07:39 <krobelus> looks like OSC 30 sets window and terminal title in Konsole.

07:40 <krobelus> iTerm and Terminal.app use OSC 1 for tab title.

07:44 <krobelus> I'll change it to only set the icon title for those two terminals

07:55 <dnkl> krobelus: thanks!

08:01 <krobelus> BTW do you think that LC_TERMINAL (and LC_TERMINAL_VERSION) make sense? The idea is that due to the LC_* prefix, those variables are passed through by SSH, so the remote program can apply workarounds for the local terminal. I think so far it's only iTerm that defines LC_TERMINAL. I guess most terminals don't need too many workarounds..

08:05 <bookworm> applications should really just check for features and not user agents....

08:05 <krobelus> true

08:06 <dnkl> yeah, I'm not too fond of setting environment variables...

08:07 <dnkl> we explicitly clear some set by other terminals just to avoid being confused with them (if foot is started from one of them)

08:07 <dnkl> if you have to know the terminal, it's better to query it with e.g XTVERSION

08:12 <krobelus> yeah that's better, and should work over ssh

11:19 <krobelus> dnkl: https://x.org/releases/X11R7.6/doc/xorg-docs/specs/ICCCM/icccm.html on WM_ICON_NAME seems to imply that the icon name is text that is supposed to be displayed, right? Not the path to a icon image. I'm not sure

11:22 <dnkl> krobelus: that sounds right. I was reading a different description of it before, that wasn't as clear as the specification itself.

11:25 <krobelus> so _NET_WM_ICON would be the path to the image

11:26 <krobelus> so it seems like an okay default to set _NET_WM_ICON_NAME to the same as the window name. But as that Xorg page says

11:26 <krobelus> > For obvious geometric reasons, fewer characters will normally be visible in WM_ICON_NAME than WM_NAME.

11:28 <dnkl> _NET_WM_ICON is an array of pimapd, if I'm not mistaken

11:29 <dnkl> I can't seem to find a way to pass an icon string under X...

11:29 <dnkl> and no escape sequence in XTerm to specify one either

11:30 <krobelus> if there is no concrete need to not set the icon name in fish I'll leave things as they are (of course setting it to the same thing as the window name is ugly)

11:31 <krobelus> I assume the icon needs to go through the terminal? The application cannot just talk to the compositor right

11:31 <dnkl> you can leave it as it is. I'll obviously have to revise my plans for xdg-toplevel-icon

11:31 <dnkl> a terminal application _can_ talk to the compositor

11:32 <dnkl> but most applications stick to the terminal interface

11:32 <dnkl> and it won't work over ssh etc