Open gauntface opened 8 years ago
yeah tool time is we relative to the system always --- so doing a 1:1 comparison between times for windows / linux will not say much given the different system stats
If i reworked the stat to be relative to each other in a column per platform... that would be interesting... :)
gaunt and I spent some time trying to get appveyor to run the build script, but windows is very unfriendly to a lot of this.
we got the basics working, though it'll require a fairly custom build file for windows. matt has some stuff moving on this.
For current status.
Irish got something working along the lines of:
Measure-Command{babel ../src/src --presets ./node_modules/babel-preset-es2015 --out-dir=dist; browserify dist/app.js | ./node_modules/.bin/uglifyjs --compress --mangle > ../src/dist/bundle.js}
I'm probably going to try and wrap it in a node script just to run each tool.
just curious, the goal here --- is it just to get windows compile time?
(and yeah the makefile approach will 100% fail on windows.. as will most of the path stuff that is hardcoded)
Another thing to remember is the time number is going to be relative to the system stats...
I did not really use the time per tool as an argument in my analysis because I knew this was going to be impossible to reproduce on other systems. if we wanted to actually use these numbers we would want to run the tools 100% on CI and dump the numbers from there
I wonder if there's any noticeable differences between Node 5 vs Node 4.
The actual time isn't important, as you said it's relative.
The main curiosity is whether some tools are faster on windows compared OS X or Linux. Lost cause?
once we have the build pipeline working on windows, we'll be able to test on the same hardware. but getting some numbers is the first step towards refining the conclusion.
I just set up tests for Web Starter Kit on Appveyor and was pretty shocked at the difference in build time.
Rollup + Babel on my linux desktop is 3s. Rollup + Babel on my Windows Surface is 30s Rollup + Babel on Appveyor is 13s
Would be interesting to see if the results of tooling change based on platform.