1 |
On Fri, Dec 16, 2005 at 12:09:10PM +0900, Jason Stubbs wrote: |
2 |
> On Thursday 15 December 2005 20:06, Brian Harring wrote: |
3 |
> > On Thu, Dec 01, 2005 at 07:41:22PM -0600, Brian Harring wrote: |
4 |
> > <snip comments on parent emerge not killing child emerge> |
5 |
> > |
6 |
> > > Either way, here's the issue, atexit registers work fine across forks, |
7 |
> > > portage.portagexit is registered prior to portage_exec.cleanup, so the |
8 |
> > > main portage pid sits there and waits for the kids to go away on their |
9 |
> > > own, then portage_exec.cleanup tries waxing the pids. |
10 |
> |
11 |
> Ok.. After a few reads, I finally figured out what you're saying here. |
12 |
|
13 |
Yeah... email to ml was just a quick update from what I tracked down |
14 |
in irc, didn't realize the email was damn near unreadable till a while |
15 |
after it was sent (and no one had commented). |
16 |
|
17 |
> > This is the only blocker for merging parallel-fetch as far as I can |
18 |
> > tell- so... my vote is nuking the wait out of portage.portageexit |
19 |
> > (it's exec subsystem crap, belong in exec's atexit (where it exists)). |
20 |
> > Assuming no complaints there, issues with parallel-fetch going into |
21 |
> > svn? |
22 |
> |
23 |
> Nuking the wait is fine if things will still work the same.. I'm kind of |
24 |
> wondering if ctrl-c'ing behaviour will change - how gets the ctrl-c first? |
25 |
|
26 |
ctrl-c doesn't come in play here- just triggers exithandler, which |
27 |
calls portageexit. What's stupid about this code is that exithandler |
28 |
has it's *own* set of kill calls in it (nothing like 101 duplicated |
29 |
bits). |
30 |
|
31 |
Either way... exit, or nothing left to execute, python will then |
32 |
trigger atexit registered callbacks in a lifo manner. So all |
33 |
exithandler really _should_ have to do is just sys.exit. |
34 |
~harring |