bash [1] is getting on my nerves at times. I could already mention how incoherent and dangerous I think the /dev/tcp/host/port trick is, but it's already known for quite some time. Debian [2] actually disables it (--disable-net-redirections), but pkgsrc doesn't [3].
Anyway, the biggest annoyance I'm facing now is with its completion, but not directly. Let me explain.
First, I'll be nice and mention I like the fact that bash' default completion is quite simple (predictable) and doesn't do too much on my behalf (unlike zsh for instance).
Now, there is something else. As far as I know, bash intercepts the SIGWINCH signal, in order to fetch information about the window size again. This is cool and works as long as bash currently receives the input from the user. When an application is running though, it seems to forward it without interpreting it for itself (I didn't study this very carefully yet).
Therefore, starting a program, resizing the window, and then coming back to the shell prompt gets you in an incoherent state, with input (completion in particular) behaving erratically until you resize the window again.
I think I saw this on Debian [4] and NetBSD [5] alike. Not sure where and how to report it though, and I don't feel like spending more time on it. So if you read this, feel free :p
[1] http://www.gnu.org/software/bash/
[2] http://packages.debian.org/sid/bash
[3] http://pkgsrc.se/shells/bash
[4] http://www.debian.org/
[5] http://www.netbsd.org/
I said I'd never blog
DeforaOS, NetBSD, reverse-engineering and stuff



