Sometimes build can fail because of an error that is not directly associated with a build step or before the associated build step ever ran(eg. if the failure happens in cache key computation). That leaves behind state where if you run buildx history inspect you get error information as expected but if you run buildx history logs then there is no obvious build step with failure.

I think we should just print the error information in the end of history logs, the same way as we do for history inspect. It may even make sense to repeat the logs for the failed step because the initial logs for that step may not be at the end of the output.