Andreas Schneider

Looks like there are some unexpected changes with libvterm 0.2. However I don't understand what those tests. Looks like something is shifted by one line sometimes.

I've implemented the buffering but this doesn't fix it.

I've built libvterm commit by commit to see what is actually breaking things. For altscreen there is actually a bugfix in libvterm (https://github.com/cryptomilk/libvterm/commit/16b857457575c3fd6ffdb0866d7a190b69c28312). However https://github.com/cryptomilk/libvterm/commit/9444c864a46e1e998299eaf7706cd6b9f75238cf seems to introduce issues instead...

There is a bug in libvterm 0.2, I have a fix for it: https://github.com/cryptomilk/libvterm/commit/cca90c80799b91203d28176cbf6f62c3e5969e7b

This looks promising. The macos-11.0 failure seems unrelated as it fails in tree-sitter ...

@clason Who could help me to understand that failing buffer_spec test better? We need to write a [libvterm test](https://github.com/cryptomilk/libvterm/blob/main/t/63screen_resize.test) for this.

@splinterofchaos Can you help?

I have a vterm test which reproduces the issue: ``` INIT WANTSTATE WANTSCREEN !neovim buffer_spec_test handles loss of focus gracefully RESET RESIZE 7,50 PUSH "tty ready\r\n" ?screen_row 0 = "tty...

The reflow patch is from the reflow repo which is a rewrite of screen reflow. I've added it to be able that more people can test it. http://bazaar.leonerd.org.uk/c/libvterm.WIP-REFLOW/ Once the...

One test which segfaults does the following: ``` RESET RESIZE 7,50 ``` Create a nvim window with 7 rows, 50 colums: Then it uses lua to create the following buffer...