On Montag, 20. Juni 2022 22:59:04 CEST Thiago Macieira wrote:
> On Monday, 20 June 2022 13:39:15 PDT Milian Wolff wrote:
> > What do I need to do to convince you to persue this patch further and
> > bring
> > vfork to QProcess? :) I'm pretty sure that this would help to reduce the
> > pain of QProcess already to such a large degree that I would not look
> > further into posix_spawn or other tricks like aspawn [1] which seems to be
> > doing clone with CLONE_VM.
> 
> You've just convinced me. I only need to clean up, add a few more comments,
> and the documentation. Oh, and test on FreeBSD, which is the other platform
> that has something similar to Linux's pidfd (pdfork).
> 
> As mentioned in the previous email, this will go into Qt 6.4 and is not a
> candidate for backporting.

Just to keep the kwrite audience in the loop:

Thiago has now submitted a fix for the performance issue here:

https://codereview.qt-project.org/c/qt/qtbase/+/417829

I guess we'll have to live with the bad performance and wait for Qt 6 which is 
somewhat close around the corner for Kate at least :)

The good news is that this only really happens on machines with lots of cores, 
when Kate is using lots of projects. A patch for the latter is already on 
invent.kde.org, so let's leave it at that.

Quite the interesting profiling story for sure, I learned a few things :)
-- 
Milian Wolff
http://milianw.de