Skip to content

Let the fork's delta merge onto upstream master - #27

Merged
archledger merged 1 commit into
plasma-fusionfrom
fix/upstream-preview-conflicts
Nov 13, 2026
Merged

archledger merged 1 commit into
plasma-fusionfrom
fix/upstream-preview-conflicts

Conversation

@archledger

@archledger archledger commented Nov 13, 2026 •

Copy link
Copy Markdown
Owner

What

The advisory upstream master preview job stops at its first step, Prepare an isolated upstream candidate: upstream.py prepare --track preview reports conflict, so no candidate is built and nothing is tested against master. The scheduled run of 2026-11-13 (43418569957, on c00da922fa) names three files. On 216857c341 the result is the same three.

Each conflict is a line of ours beside a line master changed. This moves our lines out of upstream's way.

File Conflict Change
src/tiles/tilemanager.h Our three forward declarations followed class TileModel;, which master removed when it decoupled the tile manager from the tile model (b692d1c). A namespace block of the fork's own, ahead of upstream's.
autotests/integration/tiles_test.cpp Our includes sat in upstream's blocks, one beside <QAbstractItemModelTester>, and our transient check in init() preceded the tester's line; master removed both tester lines. Upstream's include blocks and init() are as upstream has them. Our includes are in blocks of their own between them; the check is a test of its own, rootTileIsNotTransient.
src/effect/offscreenquickview.cpp The baseline set the opacity on the view's QWindow and copied it to the item. Master fixed the same bug differently (ff26dc4, "effect: Fix sync'ing of OffscreenQuickView opacity"): the item only. The fork takes upstream's two lines, so both sides make the same change.

Behaviour

Only the opacity lines change behaviour. Through the view's setter the item's opacity and what opacity() returns are as before. The view's QWindow keeps an opacity of 1 (it has no platform window, so the stored value was never used), and a value written to that QWindow directly is no longer what opacity() returns. Nothing in the tree reads the view's QWindow opacity; drawing, overlay selection and the opaque region use the item. testOpacity accepted either behaviour and pins this one now.

How long it stays clean

A probe with fourteen single upstream edits beside our lines in these three files (a synthetic commit on master each, merged with the job's own git merge-tree --merge-base=<pin>):

Clean
Lines moved just past the conflicting line (first version of this change) 7 of 14
This change 11 of 14

The three that still conflict: upstream removes class QTimer; (the line our block follows), upstream edits the #if KWIN_BUILD_X11 line our include blocks precede, and upstream edits setOpacity() again. The last one is inherent while the fork carries upstream's fix on a base that lacks it; the answer is to take upstream's new lines again.

Validation (localhost/kwin-fork-ci:fedora44)

Run Result
prepare --track preview on 216857c341 against master 75175b1045 conflict: tiles_test.cpp, offscreenquickview.cpp, tilemanager.h
The same on this commit clean, exit 0, candidate tree 7e4ff347cd (KWin 6.8.80)
prepare --track stable on this commit against Plasma/6.8 087592f44e clean
That candidate, built and run with the lane's recipe (the 18 suites) 27 of 27 test binaries
testOffscreenQuickView, testTiles on this commit 16 passed, 307 passed
testOpacity and testResynchronizeOpacity on the old opacity lines RED: three failures (the QWindow carries 0 and 0.25; opacity() returns the QWindow's 0.25)
The 18-suite campaign on this commit 27 of 27 test binaries
git clang-format --diff origin/plasma-fusion no changes

The job only runs on pull requests that touch .github/ci/, the manifest or its workflow, so it does not run here by itself. Dispatched on this branch:

Run Commit Result
43436756941 first version, b600cf1fd1 success: prepare clean, 26 of 26 suites
43441738627 this commit, 9490ad028c success: prepare clean, candidate tree 7e4ff347cd (the tree built locally), 26 of 26 suites, 2710 test rows, no failed attempt

The scheduled run reads the workflow and the source from plasma-fusion, so it turns green with the merge.

Not changed

Other places where a line of ours sits beside upstream's and merges cleanly today: fork sources inside upstream's sorted list in src/CMakeLists.txt, fork test slots after upstream's last slot in tiles_test.cpp, fork members after upstream's in tilemanager.h, and enum class Type one line from signals master added in tile.h.

Copilot AI balanced review requested due to automatic review settings November 13, 2026 13:47

Copilot AI left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota.

@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Nov 13, 2026 •

Copy link
Copy Markdown
Codex Review Summary: ✅ Completed

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-11-13T14:21:26.173267Z 9490ad0 New commits
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

The daily upstream master preview stops at its first step. The fork's
delta has three content conflicts with master, so no candidate is
built and the lane says nothing about the 397 master commits past the
pin. Each conflict is a line of ours beside a line upstream changed.

tilemanager.h: our three forward declarations followed TileModel's,
which master removed when it decoupled the tile manager from the tile
model (b692d1c). They are in a namespace block of the fork's own
now, ahead of upstream's.

tiles_test.cpp: our includes sat in upstream's blocks, one beside
QAbstractItemModelTester, and our transient check preceded the
tester's line in init(); master removed both tester lines. Upstream's
include blocks and init() are as upstream has them again. Our
includes are in blocks of their own between them, and the check is a
test of its own, rootTileIsNotTransient.

offscreenquickview.cpp: the baseline set the opacity on the view's
QWindow and copied it to the item. Master's fix of the same bug
(ff26dc4, "effect: Fix sync'ing of OffscreenQuickView opacity")
sets and reads the item only: the view has no platform window, so
the QWindow's value is never used. The fork takes upstream's two
lines. Through the view's setter the item's opacity and what
opacity() returns are as before. The QWindow keeps an opacity of 1,
and a value written to it directly is no longer what opacity()
returns. testOpacity accepted either behaviour and pins this one
now; it and the new check in testResynchronizeOpacity fail on the
old lines.

With the change the preview's merge is clean against master
75175b1.

Signed-off-by: Wisbendji Fimerlus <archledger236@gmail.com>
@archledger
archledger force-pushed the fix/upstream-preview-conflicts branch from b600cf1 to 9490ad0 Compare November 13, 2026 14:20
@archledger
archledger merged commit 0f22253 into plasma-fusion Nov 13, 2026
3 checks passed
@archledger
archledger deleted the fix/upstream-preview-conflicts branch November 13, 2026 14:57
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants