| 1 |
On Sunday 30 October 2005 02:32, Jason Stubbs wrote: |
| 2 |
> On Saturday 29 October 2005 19:34, Jason Stubbs wrote: |
| 3 |
> > On Sunday 23 October 2005 15:45, Jason Stubbs wrote: |
| 4 |
> > > Commented on the bug due to reasoning behind this patch. Essentially, |
| 5 |
> > > SIGTERM is sent to "tee", a WNOHANG waitpid() is performed followed by |
| 6 |
> > > SIGKILL if it hasn't exited. So if tee doesn't exit immediately upon |
| 7 |
> > > getting the SIGTERM, its buffers won't get a chance to get to disk due |
| 8 |
> > > to it being killed. This patch simply adds a 1 second window between |
| 9 |
> > > the SIGTERM and the SIGKILL. |
| 10 |
> > |
| 11 |
> > Per discussion on the bug, the problem is actually that tee shouldn't be |
| 12 |
> > sent a SIGTERM or SIGKILL at all. The decided solution was to wait on tee |
| 13 |
> > rather ebuild. I went a bit further than that though and cleaned up most |
| 14 |
> > of spawn(). |
| 15 |
> |
| 16 |
> Here's the full patch after incorporating feedback. It should be fairly |
| 17 |
> easy to follow given the other explanations. The only change beyond what's |
| 18 |
> already been gone over is that spawn() is now removing pids from |
| 19 |
> spawned_pids as it cleans them up. |
| 20 |
|
| 21 |
02:30.. Brain is starting to stop working. Here's a patch of _just_ |
| 22 |
portage_exec.py. With regard to splitting up patches, I agree it's much |
| 23 |
easier to parse a complete patch when familiar with the code. When one isn't |
| 24 |
familiar with the code however... Splitting it up a bit makes explaining the |
| 25 |
whys easier too - as well as keeping discussion threads seperate. |
| 26 |
|
| 27 |
-- |
| 28 |
Jason Stubbs |
Archives