There should be an optional argument for limiting parallel spawned processes

[justinwoo]

Probably using http://hackage.haskell.org/package/async-pool-0.9.0.2

Apparently, on Windows on small computers, the mapConcurrently-run spawned git processes end up with timeouts or something.

[eviefp]

Still investigating this, will update once I confirm the issue.

[eviefp]

I can confirm that changing the forConcurrently_ with a for_ and the mapConcurrently with traverse fixed the issue for us.

Any hints on how you want this solved? How about a --single-thread flag?

[kritzcreek]

with the RTS options you've already made it only spawn one thread. The single threaded runtime will still spawn multiple processes in green-threads (kind of like node does).

You really don't want to force it to only clone one repo at a time, because that takes wayyy longer. Find a good number that Appveyor can handle and limit the number of active jobs to that.

EDIT:

A quick google turned up this possible solution:

import Control.Concurrent.MSem
import Control.Concurrent.Async
import qualified Data.Traversable as T

mapPool :: T.Traversable t => Int -> (a -> IO b) -> t a -> IO (t b)
mapPool max f xs = do
    sem <- new max
    mapConcurrently (with sem . f) xs

https://stackoverflow.com/questions/18896103/can-haskells-control-concurrent-async-mapconcurrently-have-a-limit

[eviefp]

Okay, then how about --limit-threads=n?

I'll test it on our repo first to see if that works.

[justinwoo]

Fixed in #127