Skip to content

Qt: Fix HiDPI handling on X11/Windows#30399

Merged
ksunden merged 2 commits into
matplotlib:mainfrom
QuLogic:fix-qt-x-hidpi
Aug 14, 2025
Merged

Qt: Fix HiDPI handling on X11/Windows#30399
ksunden merged 2 commits into
matplotlib:mainfrom
QuLogic:fix-qt-x-hidpi

Conversation

@QuLogic

@QuLogic QuLogic commented Aug 6, 2025

Copy link
Copy Markdown
Member

PR summary

With X11, there is only ever a single scale, regardless of monitors as in Wayland, so it's always the highest scale (i.e., it's 2 if you have 1.5&1-scaled monitors). Thus we get no change events and don't update the internal scale. On Wayland though, as noted in the other issue from Qt devs, you only get the fractional scale after the first expose, so there's always at least one event there.

In the older/pre-#30345 code path, in showEvent, we'd call _update_screen to set callbacks for its changes, and that also called _update_pixel_ratio. In the new code, we don't have that initial call, and because Wayland always has at least one event at startup, it all seemed to work.

So just add the _update_pixel_ratio call in the new code path as well. On X11, this will be the highest integer scale needed and never changes. On Wayland, this will also be the rounded-up integer scale, but if using fractional scaling, will change with a subsequent event to the correct one. This does cause a few extra changes on startup, but should be more consistent across platforms.

Fixes #30386

PR checklist

With X11, there is only ever a single scale, regardless of monitors as
in Wayland, so it's always the highest scale (i.e., it's 2 if you have
1.5&1-scaled monitors). Thus we get no change events and don't update
the internal scale. On Wayland though, as noted in the other issue from
Qt devs, you only get the fractional scale after the first expose, so
there's always at least one event there.

In the older/pre-#30345 code path, in `showEvent`, we'd call
`_update_screen` to set callbacks for its changes, and that _also_
called `_update_pixel_ratio`. In the new code, we don't have that
initial call, and because Wayland always has at least one event at
startup, it all seemed to work.

So just add the `_update_pixel_ratio` call in the new code path as well.
On X11, this will be the highest integer scale needed and never changes.
On Wayland, this will also be the rounded-up integer scale, but if using
fractional scaling, will change with a subsequent event to the correct
one. This does cause a few extra changes on startup, but should be more
consistent across platforms.
@QuLogic

QuLogic commented Aug 6, 2025

Copy link
Copy Markdown
Member Author

On Wayland though, as noted in the other issue from Qt devs, you only get the fractional scale after the first expose, so there's always at least one event there.

In fact, if on Wayland you had a integer scale, then the previous code would also be incorrectly scaled, since then you don't get that first event that happens with fractional scales.

@QuLogic QuLogic added this to the v3.10.6 milestone Aug 6, 2025
@QuLogic

QuLogic commented Aug 6, 2025

Copy link
Copy Markdown
Member Author

Note, I do expect this to fix the problem on Windows as well, but I cannot test it.

@MAKOMO

MAKOMO commented Aug 8, 2025

Copy link
Copy Markdown
Contributor

Is this a X11/Windows issue only? I observe a degration on macOS as well. Window in the background is MPL 3.10.5, the one in the foreground is 3.10.3, both running on Qt/PyQt 6.9.1.

Screenshot 2025-08-08 at 14 34 33

@MAKOMO

MAKOMO commented Aug 8, 2025

Copy link
Copy Markdown
Contributor

On the positive side: adding that self._update_pixel_ratio() in showEvent() seems to fix the issue on macOS as well.

Screenshot 2025-08-08 at 14 41 57

@QuLogic

QuLogic commented Aug 8, 2025

Copy link
Copy Markdown
Member Author

It's possibly also a problem on macOS; we just haven't had any reports of such, and I can't test it myself.

@anntzer

anntzer commented Aug 11, 2025

Copy link
Copy Markdown
Contributor

I can confirm that the issue also exists on macOS, bisects to #30345, and is fixed by this PR. I'll approve on that basis (and because the fix makes sense, of course).

@ksunden ksunden merged commit 77a16b9 into matplotlib:main Aug 14, 2025
40 checks passed
meeseeksmachine pushed a commit to meeseeksmachine/matplotlib that referenced this pull request Aug 14, 2025
@QuLogic QuLogic deleted the fix-qt-x-hidpi branch August 14, 2025 18:59
tacaswell added a commit that referenced this pull request Aug 15, 2025
…399-on-v3.10.x

Backport PR #30399 on branch v3.10.x (Qt: Fix HiDPI handling on X11/Windows)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

BUG: Qt hi-dpi regression on windows and X11 with mpl 3.10.5

4 participants