01-10-2021, 05:53 PM
(01-10-2021, 03:08 PM)Tim Curtis Wrote:(01-10-2021, 02:40 PM)TheOldPresbyope Wrote:(01-10-2021, 01:25 PM)Tim Curtis Wrote: @TheOldPresbyope identified a potential problem possibly in the version of chromium-browser that came with RaspiOS 10.6 that apparently causes it to eat memory after many hours idle. I wasn't able to repro the issue on my end but that does not necessarily rule anything out in this case.
Anyway if it is some sort of issue with chromium-browser there aren't many options going forward other than to wait for a new version that may or may not contain a fix or try testing with the particular version that was used in 671 to see if the problem goes away. I don't have the bandwidth ATM to do this testing but maybe someone will volunteer to give it a try.
-Tim
Late breaking news-
A preliminary experiment suggests substituting the current Vivaldi browser (d/l directly from its devs) for Chromium as the local client may keep the system alive. I'm composing a guide for those hard souls who want to follow in my footsteps. More testers make it more likely any flaws in my approach will be uncovered.
It's not a desirable approach because Vivaldi-ARM isn't considered stable yet so isn't in the Raspberry Pi OS repo.
ETA ... but it may help the few affected users and it also may help illuminate the problem.
Regards,
Kent
Maybe more evidence it's a memory leak in chromium-browser
https://bugs.launchpad.net/ubuntu/+sourc...ug/1751137
Hadn't seen that one. If you read to the end of the thread, it appears that an ad-blocker extension was responsible in that particular case. But it was two years ago...lots of opportunities to have introduced new failure modes since. (Not that we'd know anything about that
)Chromium is an amazing mashup of code. The Vivaldi devs posted a nice blog entry about why they chose to build on the existing Chromium engine---smart as they are,"[T]here is a reason why no-one has built a new engine from scratch in 20 years."
I've now posted my guide.
Regards,
Kent
