From: Phillip Wood <phillip.wood123@gmail.com>
Date: Fri, 08 Sep 2023 13:11:51 GMT
Subject: Re: [PATCH] rebase -i: ignore signals when forking subprocesses
Message-ID: <159c16ae-4dbf-4669-bd9d-2f7c52107a68@gmail.com>
In-Reply-To: <9ba22d4b-3cbe-4d4a-8dba-bc3781e82222@gmail.com>

On 08/09/2023 10:59, Phillip Wood wrote:
>> But I think the right
>> solution is actually to start the editor in its own process group, and
>> let it be the foreground of the terminal. And then a ^C while the editor
>> is running would not only not hit git-commit, but it would not hit any
>> sequencer or other intermediate processes above it.
>>
>> I've never done it before, but from my reading we basically want to do
>> (in the forked process before we exec):
>>
>>    setsid();
>>    open("/dev/tty");
> 
> Do we want a whole new session? As I understand it to launch a 
> foreground job shells put the child in its own process group and then 
> call tcsetpgrp() to change the foreground process group of the 
> controlling terminal to that of the child. I agree that would be a 
> better way of doing things on unix.

It is better for handling SIGINT and SIGQUIT when we don't want git to 
be killed but in complicates the handling of SIGTSTP and friends. We'd 
need to pass WUNTRACED to waitpid() and then do 
"raise(WSTOPSIG(wstatus))" to propagate the signal up to the shell. When 
resuming we'd need to call tcsetpgrp() again if git is resumed in the 
foreground before sending SIGCONT to the child.

>> But of course none of that probably has any meaning on Windows. I'm not
>> sure if there are analogous concepts there we could access with
>> alternate code, or if it would need to be a totally different strategy.
> 
> Lets see if Johannes has any comments about that.

I had a quick google and it looks like cygwin somehow manages to 
implement tcsetpgrp() but the windows terminal does not have any concept 
of a foreground process group

Best Wishes

Phillip