Repository navigation
v0.1.1
Bug fixes, all of them found by running termmd on terminals its author had not:
macOS Terminal.app and iTerm2.
Fixed
- Half-block images drew a staircase of colour across the pager instead of a
picture. The backend separated its rows with a bare line feed, and the pager
runs the terminal in raw mode, where that moves down without returning to
column one. Printing hid the bug, because cooked mode supplies the carriage
return. This affected every terminal without a pixel protocol, macOS
Terminal.app included. - Chunked kitty transmissions sprayed a protocol reply into the document. Only
the first chunk carried theq=2suppression key, so a terminal that treats
each escape code independently answered the last one -- iTerm2 replies
i=0;OK-- and printing, which does not use raw mode, echoed that reply onto
the page. Only images large enough to be split showed it. - kitty images were transmitted at the default z-index, which draws them over
the text. Anything the pager drew on top of a picture -- the contents panel,
a search prompt -- disappeared behind it. They now sit below the text. - Overlay panels had no background of their own, so an image showed through the
gaps between their text. Themes gained apanelrole for it. TERM_PROGRAMnow outranks the marker environment variables when choosing a
graphics protocol.KITTY_WINDOW_IDandGHOSTTY_RESOURCES_DIRare exported
into every child process and outlive the terminal that set them, so a stale
one could send kitty escape codes to a terminal that had never heard of them.- The iTerm2 sequence carried no
name, so a receiver had nothing but a default
of "Unnamed file" to identify the payload by, and could show a file chip
rather than the image. It now sends a.pngname derived from the source, and
terminates with BEL asimgcatdoes.
Changed
- WezTerm uses the kitty protocol rather than iTerm2's, which is what the
documentation already claimed. iTerm2 3.5 speaks the kitty protocol too, and
the probe now finds and prefers it there. - The website and README explain that remote images are opt-in, which is the
first thing anyone opening a README full of badges runs into.