Script takes ~13x longer to start when comparing development version of stack with 1.9.3

[mgsloan]

Steps to reproduce

Contents of ~/tmp/hmm.hs:

#!/usr/bin/env stack
-- stack --resolver lts-13.0 script

main :: IO ()
main = putStrLn "hmm

Development version of stack:

mgsloan@treetop:~/tmp$ $(which stack) --version
Version 1.10.0, Git revision 570172b8ca0d5fd4a382abadb2951493616b0387 (6826 commits) x86_64 hpack-0.31.1
mgsloan@treetop:~/tmp$ time ./hmm.hs
hmm

real    0m3.513s
user    0m3.299s
sys     0m0.251s

I also had to wait quite a while for it to parse ~100,000 cabal files (but that is not included in this particular run)

Stack-1.9.3:

mgsloan@treetop:~/tmp$ $(which stack) --version                                                               
Version 1.9.3, Git revision 40cf7b37526b86d1676da82167ea8758a854953b (6211 commits) x86_64 hpack-0.31.1       
mgsloan@treetop:~/tmp$ time ./hmm.hs                                                                          
hmm                                                                                                           
                                                                                                              
real    0m0.281s
user    0m0.260s
sys     0m0.047s

[silky]

i think this isn't limited to just the script option; it's also slow just thinking about building, but failing:

03:53 PM noon ∈ frontend (master*) ♫ time stack build
No package revision found for: PackageName "diagrams-miso"
CallStack (from HasCallStack):
  error, called at src/Stack/Build/ConstructPlan.hs:434:42 in stack-1.10.0-HaI8Xr4wHFrAFWXvSRDn8R:Stack.Build.ConstructPlan
stack build  1.92s user 0.09s system 103% cpu 1.956 total
03:54 PM noon ∈ frontend (master*) ♫ stack --version
Version 1.10.0, Git revision 78790810649eaca7ac020d1acb9b77eab98c60ec (dirty) x86_64 hpack-0.31.1

[mgsloan]

@silky Yep, seems to be so. When building stack with itself, I see a much larger delay before the first line of output than when using 1.9.3

[qrilka]

Some of the noticed slowdown gets addressed in #4557 and #4562 improves speed of stack build.
Also #4550 should improve the situation

[mgsloan]

@qrilka Good stuff! Glad progress is being made on this

[mgsloan]

@qrilka Unfortunately those 2 PRs don't seem to have been a measurable improvement for this particular issue. I've just tested with 4e56620 and I'm getting around the same speed for running the script as before. My three runs yielded 3.740s, 3.492s, and 3.529s

[qrilka]

Now I see the problem and I'm the author actually. I will think about the ways to resolve this.

[mgsloan]

Nice, much improved! stack script still seems to have more overhead, but it's much more reasonable.

stack-2.0:

hmm                                                                                                                                  
                                                                                                                                     
real    0m0.374s
user    0m0.342s
sys     0m0.028s
mgsloan@treetop:~/oss/stack$ time ./hmm.hs                                                                                           
hmm                                                                                                                                  
                                                                                                                                     
real    0m0.356s
user    0m0.301s
sys     0m0.056s
mgsloan@treetop:~/oss/stack$ time ./hmm.hs                                                                                           
hmm                                                                                                                                  
                                                                                                                                     
real    0m0.362s
user    0m0.333s
sys     0m0.024s

stack-1.9.3:

mgsloan@treetop:~/oss/stack$ time ./hmm.hs                                                                                           
hmm                                                                                                                                  
                                                                                                                                     
real    0m0.268s
user    0m0.254s
sys     0m0.046s
mgsloan@treetop:~/oss/stack$ time ./hmm.hs                                                                                           
hmm                                                                                                                                  
                                                                                                                                     
real    0m0.258s
user    0m0.232s
sys     0m0.062s
mgsloan@treetop:~/oss/stack$ time ./hmm.hs                                                                                           
hmm                                                                                                                                  
                                                                                                                                     
real    0m0.302s
user    0m0.266s
sys     0m0.065s