Moving this to 1.1 for now as I think this both should be fixed soon and shouldn't be too hard.

I looked into this a bit and want to implement 2 things:

1. Using trap, if the release script is about to exit with an error, print a warning message to the terminal.
2. Remove all git diff calls in the release script. I initially just wanted to prevent this from exiting with an error, but I think doing that reliably in a way that still works if someone adds another call to git diff is a little difficult. Also, in the 4 years of releasing Certbot, I'm not aware of these calls ever being useful and the only keyboard input I'm aware of to make git diff exit with a nonzero status is to exit before scrolling to the end of the output.

What do you think @ohemorange? I wanted to get your +1 before spending time implementing this.

the only keyboard input I'm aware of to make git diff exit with a nonzero status is to exit before scrolling to the end of the output.

Are we sure that's what triggers the error? I'm having trouble finding where it says anything about this in the documentation; I was assuming it was an issue with taking input at the wrong point in time. Not that I have any issue with getting rid of the git diff calls; they're useful only when making updates to the release script itself.

Either way, I think using trap sounds promising. By "print a warning message to the terminal", I assume you meant "and don't exit immediately", or something? Because if it exits anyway, having a warning message printed isn't particularly helpful.

Are we sure that's what triggers the error?

I couldn't find much documentation on it, but I couldn't make it fail on any other keyboard input. git diff pipes the output to less by default. less doesn't even exit if you hit ctrl+c unless you pass -K.

Either way, I think using trap sounds promising. By "print a warning message to the terminal", I assume you meant "and don't exit immediately", or something? Because if it exits anyway, having a warning message printed isn't particularly helpful.

I am not aware of a way to catch an error at an arbitrary point in the script and then resume execution from that point unfortunately. Are you?

The idea here was just to print a loud warning like

******************************
*                            *
* THE RELEASE SCRIPT FAILED! *
*                            *
******************************

to make it clearer that it failed rather than perhaps not noticing and trying to continue with the release process.

If you don't think that's useful though I can leave it out.

I am not aware of a way to catch an error at an arbitrary point in the script and then resume execution from that point unfortunately. Are you?

I am not, but I was optimistically interpreting your comment to mean that you were, which I see now was the wrong choice (optimism).

Given that, a large message seems better than nothing. So overall, plan sgtm.

Unfortunately I don't think it's possible.

What I think we can easily do though is write better error handling in specific problem areas where the release script sometimes fails. With our current setup, I'm not aware of any after removing the calls to git diff, but if there are some, I personally think that it's worth doing and that it should be pretty easy.