13:53 <ovf> dnkl: if i enable insertion replacement mode (\e[4h), then pasting unicode stuff starts to behave weird (i.e. i get a lot of different flags when pasting a single pirate one). is this expected?

13:55 <ovf> maybe i misunderstood what this mode does, but it seems to do what i want (to behave like the normal text insertion)

13:57 <ovf> without it i'll have to reimplement insertion myself, i.e. redraw the line after the cursor

14:00 <dnkl> ovf: insertion mode is very rarely used, and not as well tested as the normal mode. There are likely bugs

14:01 <ovf> ok! i'll just reimplement it myself then. if it's rarely used in foot, it's likely rarely used elsewhere, and ultimately i want my code to work everywhere

14:09 <dnkl> ovf: if you have simple test cases (e.g printf one-liners), by all means, open an issue

14:14 <ovf> dnkl: thanks! that was a lot of form filling for that one printf line. :-) https://codeberg.org/dnkl/foot/issues/1947

14:35 <dnkl> yeah... but less ping/pong with all the standard questions. People in general are terrible bug reporters ;)

14:38 <dnkl> ovf: ah, ok, so, this particular issue is that, since all of the codepoints "append" to the previous one, the cursor is moved back to the previous cell. When the insert logic runs, everything from the cursor and on is memmove:d one cell to the right. This means the last flag-codepoint is copied to the next cell, leading to a string of emojis

14:39 <ovf> ...which is, unsurprisingly, exactly the issue my naive client-side reimplementation had. :-)

14:41 <dnkl> just to illustrate...

14:41 <dnkl> printf "xxx\b\e[4h\U1f3f4\e[4l\U200d\U2620\Ufe0f\n"

14:41 <dnkl> "works"

14:41 <dnkl> (flag is inserted before the last x)

23:26 <krobelus> so the sane way to query the terminal is to do a blocking wait (on Primary DA response). Some people are suggesting adding a timeout which seems wrong, I'd rather work around / fix the broken terminals. Not sure what other querying programs do

23:39 <krobelus> even the most plain terminals respond to Primary DA; so far I've only found Midnight Commander and dvtm that don't (to be fixed easily)

23:40 <rockorager> krobelus: I put a 1 second timeout

23:40 <rockorager> I forgot which terminal I’ve encountered that doesn’t respond…

23:41 <rockorager> notcurses has good notes on it

23:42 <rockorager> I think st?

23:46 <krobelus> I'm fine with adding such a timeout (though I've had >1s ping at times) but it should cover up the need to fix terminals. If we fix them, future users will get by without timeout.

23:48 <krobelus> rockorager: that's in libvaxis right?

23:53 <rockorager> yeah

23:53 <rockorager> I agree about fixing broken terminals

23:54 <krobelus> rockorager: where's the notcurses docs? I'm reading through some of their issues

23:54 <krobelus> ah, TERMINALS.md

23:54 <rockorager> I wonder if you the broken ones respond to DSR (Ps=5)

23:55 <rockorager> you could do a DSR and a DA1 and break on either

23:55 <krobelus> > if a terminal does *not* reply in a recognizable way to Primary Device Attributes, `notcurses_init()` will hang

23:56 <rockorager> https://github.com/dankamongmen/notcurses/blob/master/src/lib/termdesc.c

23:56 <rockorager> also in there ^^

23:56 <rockorager> which has some absolutely amazing commentary around Terminal.app 😂

23:57 <rockorager> btw I finally started using the 4.0 beta, awesome work