opencode: fix light theme

[losnappas]

Fixes light themes. Like here #2192 , but I think this is a better fix. This way stylix handles the polarity, not opencode.

Before (catppuccin-latte):

After: on the left, the opencode builtin catppuccin (light) theme. On the right, the stylix catppuccin-latte theme applied.

Some differences, but better than what was there before (+ I think these diffs exist on dark theme too).

Supersedes: #2192
Supersedes: #2322

---

• I certify that I have the right to submit this contribution under the MIT license
• Commit messages adhere to Stylix commit conventions
• Theming changes adhere to the Stylix style guide
• Changes have been tested locally
• Changes have been tested in testbeds
• Each commit in this PR is suitable for backport

[trueNAHO]

Could you elaborate on this? Is there no more reliable solution than manually inverting dark and light values, otherwise what is their purpose?

I don't have the slightest idea. Could be just broken on their end, I guess. Devin thinks that should've worked all along. I'll figure it out and get back to ya

-- #2322 (comment)

Based on the light theme seemingly working at the time of #1945 (comment), this seems like an upstream breaking change or regression. If this is a regression, it might be worth opening an upstream issue.

Some differences, but better than what was there before (+ I think these diffs exist on dark theme too).

Are you referring to commit b135edb ("opencode: brighten default text colors (#2100)")?

Fixes light themes. Like here #2192 , but I think this is a better fix.

At least this seems less hacky than the previous approaches of manually inverting polarities. Would be great to get approvals whether the new light theme looks appropriate.

• Each commit in this PR is suitable for backport

Are you sure this issue is present on the stable branch? Maybe the nixpkgs/release-26.06 branch does not have this behavior yet.

CC: @Azd325, @Juanal07, @Programmerino, @arunoruto

[Juanal07]

I refreshed stylix pointing to the last commit and is still broken but then I tried b135edb and it worked so it seems like a regresion as you said.