0.19.0 became very slow in development environment

[oeyoews]

I upgraded from 0.18.15 to 0.19.0, and I obviously felt that the compilation speed in the development environment became very slow. Does anyone have similar experience?

[aklinker1]

Does using entrypointLoader: 'jiti' fix the performance issues?

[oeyoews]

Does using entrypointLoader: 'jiti' fix the performance issues?

I haven't used this yet, I'll try it tomorrow

[aklinker1]

Some others have reported errors with 0.19.0's vite-node loader, and I've been releasing some alpha versions for them to test.

I just released [email protected] that includes a performance change to tree-shake out any code that isn't related to the entrypoint options. It should be as fast or faster than v0.18 and below.

Edit: v0.19.1 is officially out, give that a go instead.

pnpm i [email protected]

[jaek-s]

Hey! I'm jumping in here because I'm experiencing the same issue. For my extension, v0.18 will give me the "Opened browser" message in under 2s or so. After upgrading wxt to v0.19.4 will take longer than 6s, usually closer to 8s before the "Opened browser" message appears.

I tried adding entrypointLoader: 'jiti', but that didn't seem to have any effect.

I'm really not sure what could be causing the issue. Maybe some dependency I have, or just the size or complexity of my project? I noticed that there was some slowdown after upgrading some of the examples to V0.19 and downgrading github-better-line-counts to v0.18 resulted in some speed up. Nothing as dramatic as what I'm seeing, to be sure.

Let me know if there's anything I can do to help track this down!

[1natsu172]

Yes, I don't know why, but it start up faster the second time around. Something is still lurking. Let's keep
investigating.

@jaek-s BTW, is your machine MacOS or Windows?

[jaek-s]

I'm on MacOS with an M1 chip

[1natsu172]

Thank you. I add it to the sample of information 👍

[1natsu172]

Alright, based on Aaron's bench, I figured out why the performance started to slow down.

#866 (reply in thread)

• 48783ba
• 5eb5be7 (or f1bb7cb?)
• And maybe a0971ca?

It was caused by 48783ba commit that bumped the vite 5.2.8 to 5.3.0. There was a change vitejs/vite#16453 in 5.2.9 of CHANGELOG that caused this. The change was to make vite.watcher not ignore outDir when emptyOutDir is false. So until 5.2.9, outDir was ignored and performance was stable.

I have not yet figured out why the slow down accelerated with 5eb5be7 (or f1bb7cb?) commits. If something is added that increases the output to outDir or .wxt, If not, there must be a cause other than vite.watcher.

[1natsu172]

I looked closer, the commit that started spiking is 14011c1 . Hmmm……that vite-node…

[1natsu172]

On my machine, I was doing some testing and found that even with firefox, occur the second launch faster issue.

In addition, launching pure chrome with chrome-launcher without WXT was almost the same as the [email protected] launch speed. And also had the second launch faster issue.

If everyone else is the same, then this second faster problem might be a macOS issue. I also found the interesting thread. (https://www.reddit.com/r/MacOS/comments/1bp7ugd/why_does_it_take_so_long_to_open_apps_for_the/)