Thanks for fixing this! I usually bootstrap my OS X installation with relevant gnu tools and replaces the bsd tools, so I didn’t catch this one :/

What is the performance of shasum (perl) compared to sha1sum (c)? If sha1sum is much faster, we could update the README with the brew install step.

To test the performance you could run the command on the entire src folder in tmrow. :)

Thanks for fixing this! I usually bootstrap my OS X installation with relevant gnu tools and replaces the bsd tools, so I didn’t catch this one :/

What is the performance of shasum (perl) compared to sha1sum (c)? If sha1sum is much faster, we could update the README with the brew install step.

To test the performance you could run the command on the entire src folder in tmrow. :)

I tried removing everything except the hashing to actually test in on Drone, and there is a significant difference - now sure if it's enough to ask all brick users to also run a brew install md5sha1sum first though :)

Thanks for testing the performance so carefully! I thought you would just test locally.

Not sure how much it affects the overall performance? It is just a few seconds, then let us not bother our dear users with a brew install step.

If it is more substantial then I think an additional install step is acceptable for an increase in build time.

Thanks for testing the performance so carefully! I thought you would just test locally.

I also tested locally, but I've come to the conclusion that everything takes way longer when running in CI - so I don't really trust local comparisons for this 😄

Not sure how much it affects the overall performance? It is just a few seconds, then let us not bother our dear users with a brew install step.

If it is more substantial then I think an additional install step is acceptable for an increase in build time.

The screenshots shows running only the get_hash function for all packages, so it's 1.8s for sha1sum and 5.53s for shasum. So I think we should stick with shasum for now as it's a fairly small costs for users unless they have listed a ton of inputs (which they probably shouldn't anyway).