You are right. It it doesn't account for there being only variable fonts.
I don't think we will work well in such a case - meaning lots of our
regression tests would fail.
But it would be better to have at least some font available.
https://bugs.openjdk.org/browse/JDK-8397728
-phil.
On 11/17/26 10:21 PM, Tianon Gravi wrote:
On Tue, 10 Nov 2026 18:12:22 GMT, Phil Race <[email protected]> wrote:
This fix resolves problems with styles of logical fonts on OL 10.2 - and any
other Linux where a font variation is used as the first component of the Linux
sans/serif/monospace desktop fonts.
It skips such initial fonts - as identified by [wght] in the name, which seems
sufficient for now, although not fool proof.
The alternative is opening the fonts to see if they are variation fonts during
the process of gathering the names.
That is something we don't really want to do. That's why we use fontconfig.
If we need to refine this later, we will, but hopefully before that is needed
we'll have proper variation support.
---------
- [x] I confirm that I make this contribution in accordance with the [OpenJDK
Interim AI Policy](https://openjdk.org/legal/ai).
Phil Race has updated the pull request incrementally with one additional commit
since the last revision:
8396849
FWIW, this causes the font selection code to fail hard if nothing _but_
variable fonts are installed (such as in a minimal Docker container), which I
don't think was the intent 😅
(https://github.com/docker-library/openjdk/pull/559)
-------------
PR Comment: https://git.openjdk.org/jdk/pull/33761#issuecomment-6970839609