hang on exit - bug or no bug?

[Phil Howard]

Nicolas Williams wrote:

> This patch seems to do the same sort of thing being discussed in the
> SIGCHLD race thread: add a timeout to the select(). If this is done only
> when there's no child fildes left open, and only to break the SIGCHLD
> race deadlock, then it can't lead to data loss (unless you're doing
> something else wrong).

Actually, I would like to see a way to solve the situation where there
are hung processes.  In an interactive session I can do ^\. and get out
OK.  In a "batch" session this normally should not be a problem, but I
have seen it occur.  Now I do not want to trade it off to get the data
loss problem as that is worse.  If both cannot be fixed at the same time,
then hang is preferrable to data loss.  But I do think it should be
possible to get around the problem without data loss (aside from the
ambiguity of dealing with background processes, which for batch scripts
can be done the "right way" anyway).

I'm following this thread not because I have an urgent problem with
this (I don't) but rather to just get an understanding on the group
think and philosophy on how things like this are dealt with.

-- 
-----------------------------------------------------------------
| Phil Howard - KA9WGN |   Dallas   | http://linuxhomepage.com/ |
| phil-nospam at ipal.net | Texas, USA | http://phil.ipal.org/     |
-----------------------------------------------------------------