Krzysztof Kotowicz

I traced where the spurious diff changes are coming from. Bikeshed generates IDs for term references for the definition panels using "circled digits" (`id="ref-for-trusted-types-sink-group②`). The circled digits are just sequential...

> [setting script content via innerText] could lead to some script execution errors despite being set via a trusted object Isn't this the case even without Trusted Types? The spec...

> Specifically I mean trusted types related script execution errors. Because the value may mismatch in which case TT will reject it, which could be unexpected. I might be missing...

Again, I don't think the TT claim that the values reaching the sinks are semantically correct, only that they were checked before reaching the sinks.

The current export output color profile is also stored in a DT preference (`plugins/lighttable/export/icctype=26` for `Display P3 RGB`), but I can't read core DT preferences from Lua (only enumerate them...

> Which leaves option 1 (at least for now)... That's what I did in the end in https://github.com/darktable-org/lua-scripts/pull/502/commits/6a835ce1a271439d2971756eb6fb95e478ebab67, with a slightly improved method. I go to `first` option, and then...

FWIW this workaround still has occasional race conditions, I noticed I had to *increase* the sleep time when switching to faster device recently.

Thanks, I like using the offset trick! So it seems like we either have the race condition or hardcoding the profile display order (and TBH the latter seems preferred now)....

For context, these were removed from the spec in https://github.com/w3c/trusted-types/issues/305. @otherdaniel, we should correct the Chromium impl and remove the enforcement.

@otherdaniel how does Sanitizer deal with this? From memory, the challenge was that there was no centralized and specced list of event handlers we could refer to.