search

hishamhm / htop

htop is an interactive text-mode process viewer for Unix systems. It aims to be a better 'top'.
GNU General Public License v2.0
5.83k stars 581 forks source link

Fix the error that the period variable obtained for the first time is abnormal #1024

Closed NachoLulu closed 4 years ago

NachoLulu commented 4 years ago

Hello, when I was watching and learning the source code, I found that the period variable data obtained for the first time was abnormal, so I judged the variable data of this-> cpus[0].totalPeriod,Fix the error that the period variable obtained for the first time is abnormal.

natoscott commented 4 years ago

This change looks like a workaround rather than a root cause fix? I guess the question is, why the 'period variable obtained for the first time is abnormal'... any idea?

I wonder if the rapid-fire double sampling done in htop.c main() is part of the story here:

   ProcessList_scan(pl);
   millisleep(75);
   ProcessList_scan(pl);

could we end up with an abnormal period value due to this perhaps? (we need this code so that counters can be rate-converted) It is using a (short) internal (75msec) different to the regular sample interval - maybe a calculation somewhere uses the wrong time interval? sorry, bit of a wild guess :) - but we do need to understand the underlying problem properly still I think.

NachoLulu commented 4 years ago

Because when I was using htop, I found that the CPU utilization in the first second after opening the program was displayed as 100%. When I tried to solve this problem, I found that the output variable period was abnormal, and only the first abnormally large , So I first wanted to rule out this situation, so I made this change.

NachoLulu commented 4 years ago

I thought about your guess carefully, I think it makes sense, I will confirm it in the direction of your guess, thank you for your reply

natoscott commented 4 years ago

I thought about your guess carefully, I think it makes sense, I will confirm it in the direction of your guess, thank you for your reply

@Nacho-meteor no problem at all.

There definitely seems to be an underlying issue here somewhere - I've seen quite a few people mention it now; keep at it, I'm sure you'll figure it out.