(no subject)

[torkell]

It would be nice if CVS prompted before overwriting a large directory tree with an older branch.

Scratch that, it would be nice if unix stuff in general prompted before doing file-trashing operations.

Comments

[tau-iota-mu-c.livejournal.com]

I hate cvs commit -m -- because if I accidentally hit the enter key instead of "'", then I end up committing the entire tree (often prematurely) with a bogus commit message.

[ralesk.livejournal.com]

Doesn’t having an EDITOR set help, like in SVN’s case? Then you don’t have to do -m at all, it’ll just pop up an editor with the default commit text.

[tau-iota-mu-c.livejournal.com]

Hmmm. My editor is slightly heavy weight, and sometimes I want a really quick commit from the commandline (and besides, cvs commit by itself will still act as a wildcard - just hopefully you have an $EDITOR that doesn't trap SIGINT, so you can kill the cvs process).

[ralesk.livejournal.com]

Many file-related tools have a -i option for interactivity, if I remember correctly. Would be nice if that was the default and you could turn it off for scripting (like “-y, assume yes for every question” or something).

[tau-iota-mu-c.livejournal.com]

Most linux people define aliases to mv/cp/rm -i. I find this dangerous. Log into a system that is not set up by you, type rm expecting confirmation, and oops, you just blew away *.

I prefer to define aliases called rmi/cpi/mvi, and my fingers tyop them instead. And on my own systems, rm/cp/mv became aliases that echoed "look, ya silly bugger. Use rmi". After a few months of that, now I type rmi without thinking. Scripts still found the real rm/cp/mv. If I logged into someone elses system, rmi would come back with "command not found". Then I would go back and edit the command to do rm -i. Failsafe. Not dependant upon quirky ways other people might set their machines up. Safe. I like safe.