On 6/1/2023 12:16 PM, Richard Henderson wrote:
> On 5/31/23 20:19, Wu, Fei wrote:
>> On 6/1/2023 8:01 AM, Richard Henderson wrote:
>>> On 5/30/23 01:35, Fei Wu wrote:
>>>> +/* TBStatistic collection controls */
>>>> +enum TBStatsStatus {
>>>> + TB_STATS_DISABLED = 0,
>>>> + TB_STATS_RUNNING,
>>>> + TB_STATS_PAUSED,
>>>> + TB_STATS_STOPPED
>>>> +};
>>>
>>> I don't see what PAUSED or STOPPED actually do.
>>> As far as I can see, stats are either being collected or not: a boolean.
>>>
>> If STOPPED, clean_tbstats() gets called, all the tbstats history is
>> destroyed, but it's not for PAUSED.
>
> But it doesn't PAUSE everything either, since (1) code is built into
> each tb, and (2) each tb->tb_stats->stats_enabled neither changed.
> Indeed, tb_stats_collection_enabled() is only checked in a fraction of
> the places stats are collected.
>
tbs are always flushed at the end of hmp tb-stats cmd by tb_flush(),
stats_enabled can be changed e.g. during pause/filter cmd. It should be
a bug if stats are collected w/o tb_stats_collection_enabled().
Thanks,
Fei.
>
> r~
>
