[Yum] Why is yum not interruptible with SIGTERM/ctrl-c?

[David Timms]

Michael Stenner wrote:
> On Thu, May 04, 2006 at 11:20:41AM -0400, seth vidal wrote:
>> On Thu, 2006-05-04 at 12:04 -0300, Juergen Botz wrote:
>>> Maybe this is a stupid question... and to be honest I have not
>>> looked at the code.  But there have been many times that I 
>>> really needed to terminate a running yum, and right now I have
>>> to use SIGKILL and then I usually need do a 'clean metadata'.
>>> At best this seems to be bad UI... users need a way to cleanly
>>> stop any process on the system.
>> in the majority of the situation it is due to rpm grabbing those signals
>> so that a user will not cancel a process in the middle of its database
>> being written to which could leave the system in a bad state.
> 
> I recall that some folks (a little myself, but mostly Menno, I think)
> put a bunch of time into this problem and came up with some clever
> solutions.  I think the plan was that during a download, one ctrl-c
> would skip to the next mirror and two would bail out of yum.  Did that
> code ever make it in?
That code is present (for some time in FC5 and late FC4):
one ctrl-c skips to next mirror
maybe 5 or 10 ctrl-c or held down (5secs) will can the whole process 
unless performing the actual rpm transaction part.

I guess the test transaction part should be nuke-able, because yum/rpm 
is only going through the motions.

Pup is also killable with the X / wait 5 seconds / end process option 
Although the GUI goes away at end process, I think the yum process 
continues until the current download completes before exiting. (ps aux)

DaveT.