Re: KDevelop continuous activity?
René J.V. Bertin <[email protected]> Thu, 09 Apr 2020 19:25:04 +0200
| Newsgroups | gmane.comp.kde.users.kdevelop |
|---|---|
| Message-ID | <6615223.Fk9BJN0lTm@bola> |
On Thursday April 09 2020 16:51:14 Milian Wolff wrote: >perf record --call-graph dwarf -p $(pidof KDevelop) >hotspot perf.data Thanks! I had a hunch it would be using perf but got a bit lost in its documentation. >Drawing the cursor in the text editor? Thankfully, no. I did see something from the KTextEditor library show up in the samples on Mac, but not a high value and it's probably not representative. The cursor animation is stopped when the window loses focus, so I'd hope that also stops CPU usage. The real culprit appears to be the glib-based event processing, which for some reason must be falling back to a polling approach for me on both platforms. I'm quite certain that's new; the TCL-only session I mentioned is usually open for days, so the total time value becomes ... noteworthy. Now, did I in fact see the kind of very long total_time values but didn't register them because, "yeah, that app has indeed been open for days"?? Anyway, other than checking for (or processing) events the process seems to spend a lot of time creating the DUChain cleanup thread (surprising, in a session that only has TCL and patch files open?). And in some libc function called by qt_format_text. I didn't see these 2 culprits on Mac. If you want I can send you the perf.data file, but I don't know if there's much you can do with it, without having all my binaries too? Cheers, R.
Screen Shot 2020-04-09 at 19.03.15.png
(image/png, 631.1 KB) - not displayed