Great, it sounds like the recent changes are helping. Thanks for the confirmation.
I’ve hinted at Tim that it’d be time to do a release. (Luke is the TinyFPGA creator, but hasn’t had any time to do any tinyprog/bootloader related work in at least a year, which is why Tim persuaded Luke to allow Tim to do tinyprog uploads to PyPI.)
FTR, one of the main reasons the new version is a lot faster is that one of Tramell Hudson’s changes avoids even erasing 4k blocks where the version already in the SPI flash is identical to the version that is in the file being uploaded. So for unchanged/hardly changed files the “upload” turns into basically just reading the SPI flash (to check it’s the same). Definitely should help with quick turn around on iterative development.
PS: There’s also an outstanding issue – https://github.com/tinyfpga/TinyFPGA-Bootloader/issues/55 – to create another retry layer (based on an old, unmerged, pull request) which hopefully will solve 95%+ of the remaining problems (since aside from exact-size related issues, the remainder seem to be “works if you run it again” type issues).