Re: [PATCH] taskstats: retain dead thread stats in TGID queries

[Yiyang Chen]

Hi Yaxin,

Thanks for the review.

> 1. Capturing the final delay-accounting update before task exit is necessary;
> otherwise, cumulative task delay will be underestimated over time.

Yes, exactly. My patch preserves this behavior — delay accounting was already
being accumulated in fill_tgid_exit() via delayacct_add_tsk(). What I'm doing
is extending this same pattern to other fields (ac_etime, ac_utime, ac_stime,
nvcsw, nivcsw) that were only accumulated for live threads but not for exited
threads, causing data loss in TGID queries.

> 2. For fixes of this kind, it is best to include the issue/bug ID that is being fixed,
> so the change remains clearly traceable.

Thanks. There is no formal issue/bug ID for this problem at the moment.
If needed, I can try to report one and update the patch accordingly.

Thanks,
Yiyang Chen