Can we make network tasks more resilient?

[jzmaddock]

This is just me thinking out loud here, but tasks occasionally fail due to network issues: https://drone.cpp.al/boostorg/multiprecision/644/147/1, for network tasks within boost.sh I added some simple resilience like this:

for i in 1 2 3 4 5; do git clone -b $TRAVIS_BRANCH --depth 1 https://github.com/boostorg/boost.git boost-root && break || sleep 15; done

There may well be better ways of doing this frankly, but what I'm missing is some way to do something similar for the initial clone step that happens before any user-scripts run?

[sdarwin]

Fixed.

You linked to an s390x job. Retries for the initial git clone step were added in multiple parts during the last year. The standard linux runner was working, but the s390x runner hadn't been updated. I have refreshed those machines now.

What's still outstanding are Windows, macOS. However those are difficult because Harness (the drone company) has not implemented retries there.

[jzmaddock]

Wow, that's quick, many thanks as ever Sam!