Repository navigation
untested HUB75 driver used in 16.0.0 release #5594
Description
Lines 702 to 703 in 560eb37
#5558 has upgraded the HUB75 driver version, unfortunately picking a very new, very buggy release of the driver 😢 . The previous library version we used was https://github.com/mrfaptastic/ESP32-HUB75-MatrixPanel-DMA.git#3.0.11. Even with some known stability issues, it was tested for a longer time and kind of qualified for WLED as proven-in-use.
While the upgrade from 3.0.11 to 3.0.14 was mentioned in the PR description, I've obviously overlooked this change, so I missed my chance to veto.
Unfortunately all HUB75 driver releases since middle of last year are full of bugs - for example when running certain driver chips at full brightness, the library will produce skewed colors and other ugly artefacts. For bitdepth != 8, panel display is very dark, and other crazy issued that surfaced during my testing in WLED-MM.
Lessons learned: PLEASE do not upgrade a known-fragile driver library in the last minute before a WLED release.
I'll need some time to identify a well-working driver library commit hash.
Activity
netmindz commentedon May 11, 2026
I had run on all 4 hardware setups and did soak test for hours and I did check the release included the hotfix we were using for S3 when I chose what version to standardise on to be used for all builds.
I was unaware you had tried and failed with newer builds sorry. The hub75 builds were missing from the beta release entirely so we missed the opportunity to shake out any issue for that
[-]buggy HUB75 driver used in 16.0.0 release[/-][+]untested HUB75 driver used in 16.0.0 release[/+]softhack007 commentedon May 12, 2026
@netmindz sorry, I think the "quality manager" in me was freaking out about "WTF why replace a driver 5 minutes before flight".
Actually, my bad experiences come from version 3.0.12 / 3.0.13, and a few intermediates after 3.0.13.
I've just tried to reproduce some known issues using release 3.0.14, and it is better than expected - reducing color depth does not reduce brightness anymore. Also, I could not (yet) reproduce false colors in high brightness.
It was still a dangerous thing to change the library with just a bit of testing. But it looks like I can reduce the warning level 😃
I'll continue testing with release 3.0.14.
netmindz commentedon May 12, 2026
Yeah it was perhaps a bit hasty, I just didn't like adding the tech debt of different lib deps for different esp32 builds if a hotfix had been merged, but didn't properly consider other changes that might be in that release