This was just supposed to be a new feature for the generic image tool (dithering support), but while working on that things kind of snowballed 😅
* Added dithering support to the generic image tool and the boot logo changer, when converting images to RGB565 format. This uses a Bayer 8x8 matrix, and the overall "strength" of the dither can be controlled - it defaults to what I feel is a sane value. Dithering can help reduce banding effects due to the low colour depth in RGB565.
* Made "fix" scaling mode more flexible in the generic image tool - now there's checkboxes beside the width and height dimensions - if you un-check one, the other dimension will be calculated automatically to keep the input's aspect ratio intact
* Improved downscaling quality in the generic image tool. While working on the dithering feature, I discovered the previous "gaussian resampling" downscaling method I was using introduced distortion in certain situations. I had a lot of fun playing around with possible replacements (I tried 10 new downscaling functions!), and finally settled on a hybrid function that mixes powers-of-two downscaling (with some custom mipmap style cross-blending) with Hermite interpolation. This new method is reasonably quick, gives clean results with no great distortion, and works well with alpha channels (doesn't introduce any dark fringing)
* Generic image tool now shows the dimensions of the output image
* Added a max width on the main tool page bodies, so that they don't get so wide on full-screen desktop browsers
* Fixed a few edge-case logic bugs here and there (e.g., when upscaling only a single dimensions with the generic image tool, or places where I thought I was copying objects, but was only creating references to them, etc.)
* Switched from using "var" declarations across all tool codebases to "let" or "const" instead
* Switched out the SVG alert icons to just use emoji instead
* Various other nips and tucks (fixed up my arbitrary 80-column comment wrapping on the tools that had previously just been eye-balled, fixed a few comment typos, etc.)
Looks like there's a new official firmware version out, and it's got a critical bug with SNES save states - avoid this one!
Updated the tools I host to support the new firmware, complete with warnings about it. Also updated the main Github document with the first details about the firmware, more will come later when my own download of it has completed.
Just a tweak really - re-ordered the zeroing-out checks in getFirmwareHash() to match the expected order of each section within the BIOS file. This may help to avoid a theoretical issue where bytes in user-substituted content (e.g., a boot logo) accidentally matches against some other blob of binary we're searching for using findSequence(). Doing the search-and-replace in the expected ordering should help to avoid that... I think? Won't hurt anyway!
Added support for checking firmwares that have modified the two byte-pairs `bnister` identified as a workaround for the whole "start-SNES-games-twice" issue that cropped up in firmware versions after March
Some recent work by `dteyn` and `bnister` now allows for the bytes responsible for power monitoring on the SF2000 to be changed to more sensible values. I've updated the BIOS version checking details to account for these bytes, so that `bisrv.asd` file versions can still be detected correctly even if the user has altered their power monitoring bytes
There's a new BIOS revision floating around (MD5 hash first posted in Discord on August 3rd); no official announcement from Data Frog about it yet, but I've gone ahead and added support for it to the Boot Logo Changer and the Button Mapping Changer tools
There was a lot of repeated JS across all my tools (e.g., image conversion, BIOS hash checking, etc.), and so I decided to move all shared code out to its own separate library. This will improve scalability and maintainability at the cost of portability - a fair trade in my opinion.
This work also involved large-scale refactoring of all of the tools. There's no actual functional difference (all the tools should behave exactly the same), bar some UI improvements for the oldest tools as a result of moving to the same codebase used by the more recent tools.
