So I have a a custom triplet with per port customization and this is a problem for the binary cache since if I add a new customization everything will be invalidated, removed and rebuild although only one port was effectively changed.

I thus suggest: Instead of using the file hash of the triplet to determine the hash for the binary cache it would be better to get the variables set by the triplet and use those to define the triplet hash for a port. Since the triplet is already read numerous times for compiler hash checks it shouldn't add a lot of additional overhead.

This would greatly improve my workflow if I have to touch the triplet file.