D7995: KDevelop: address dirwatching inefficiency (WIP/PoC)

[René J.V. Bertin]

rjvbb added a comment.


  >   I care, otherwise I wouldn't review. Simplify the code, then we can run the benchmark and see what it brings.
  
  I already know what the benchmark brings, and what's more, I know what my implementation brings in actual KDevelop sessions (the good and the bad).
  
  I have tried to explain several times why I think the ProjectWatcher class is justified and why I don't want to use KDirWatch directly. We'll see what feedback I get on my question about the underlying main reason but as long as I'm not convinced KDW::addDir() can be called directly we'll remain in the current stalemate.
  
  Simplifying? I've been doing the opposite since I gave up on upstreaming. I've been adding the simple measures against multiple concurrent reloads I posted elsewhere (too much hassle to maintain that as a separate patch). And a way to opt out and use the old dirwatching mechanism which does perform better for certain kinds of projects even if they take longer to load. That doesn't really make the patch more more complicated of course, and it should make benchmarking old vs. new more trivial (no more rebuilding with or without the patch).

REPOSITORY
  R32 KDevelop

REVISION DETAIL
  https://phabricator.kde.org/D7995

To: rjvbb, #kdevelop, mwolff
Cc: aaronpuchert, arrowdodger, kfunk, dfaure, mwolff, brauch, kdevelop-devel, njensen, geetamc, Pilzschaf, akshaydeo, surgenight
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kdevelop-devel/attachments/20171125/3f651e3a/attachment.html>