Johannes Altmanninger

a deterministic reproducer would be great. It might be fixed in latest master but we don't know yet. With a reproducer that's easy to test

This is fixed since the new logic keeps `self.event_storage` always empty (only the heap allocation is reused but only after being swapped into a local variable). If it is decided...

There are various other uses of `env!`, like `const DOC_DIR: &str = env!("DOCDIR");`, I wonder if they are affected too / why not. If they are, they should crash too...

is there a particular reason you don't want to patch this downstream? E.g. the minimal fix; for upstream I'd rather have a complete fix. But it seems a bit hacky....

I haven't had time to study this so take it with a grain of salt; my feeling is that we prefer not adding something that we'd eventually want to get...

> The issue is that it's pretty easy to break (move the fish binary, I think `make install` has the same problem. We probably already print warnings on startup if...

I can reproduce by using `C` to create a commit that uses the external editor. Simpler repro: fish -C 'bind ctrl-g bash repaint' bash works but fish fails to re-enable...

> `echo -e 'prompt \e[48;2;242;229;188m\e[K\nfoo\nbar\e[49maaaaa\nbaz '` Maybe it was 263f1b35de (Reapply "Clear to eol before outputting line in multi-line prompt", 2024-09-28)? Though I don't immediately see how..

probably fish should first erase all lines and then draw the prompt verbatim

FWIW I want to remove a workaround for dvtm/abduco from fish, in order to make things simpler and more predictable. The workaround is that fish doesn't try to do query...