snapicon.app, cpfg.app, and medit.app share an identical CFBundleIdentifier on macOS

[JimHanan]

On macOS (Apple Silicon, native arm64 build from source), snapicon fails
to capture the correct window content — it captures the desktop
background/wallpaper instead of the window beneath the capture frame, and
in one attempt to work around this, it quit unexpectedly instead.

While investigating, found that three different plugin bundles all share
the literal same CFBundleIdentifier, apparently a copy-paste artifact:

snapicon/Info.plist: org.algorithmicbotany.CPFG
cpfg/Info.plist: org.algorithmicbotany.CPFG
medit/Info.plist: org.algorithmicbotany.CPFG

(browser/Info.plist correctly uses org.algorithmicbotany.Browser, for
comparison — so the pattern elsewhere in the tree is one identifier per
tool.)

On modern macOS, Screen Recording access (needed for snapicon's
CGWindowListCreateImage-based capture, via Qt's QScreen::grabWindow) is
authorized per bundle identifier via TCC. Three unrelated tools claiming
the same identifier seems likely to confuse that permission tracking —
consistent with, though not conclusively proven to be the sole cause of,
the capture-background-instead-of-window symptom.

Suggest giving each plugin bundle its own unique CFBundleIdentifier
(e.g. org.algorithmicbotany.Snapicon, org.algorithmicbotany.Medit),
matching how browser.app is already set up.

[mikcieslak]

I will update the CFBundleIdentifier, but I'm not sure if that is the issue because snapicon works without issues for me. Did you compile the main branch or the migrate/qt6 branch?

[JimHanan]

According to Claude:
main — not migrate/qt6. Four independent lines of evidence agree:

Current branch is main, at 82bff3b ("Merge pull request #13 from gilgamec/main", 2026-06-10).
The reflog has exactly one entry — the original clone on 2026-07-08. HEAD has never moved, so no other branch was ever checked out.
The clone is single-branch and shallow. Its fetch refspec is +refs/heads/main:refs/remotes/origin/main, so only main was ever fetched — migrate/qt6 isn't even present locally to check out. (I tried to query the remote for its real branch list, but the network call returned nothing, so I can't confirm from here whether that branch exists upstream.)
The built binary links Qt 5.15.19 — QtCore/QtGui/QtWidgets/QtOpenGL all Versions/5. A Qt6 migration branch would link Qt6 frameworks.
browser.app's binary is dated 2026-07-08 16:42, the same day as the clone, consistent with it being built from this tree.

[mikcieslak]

I tested the released version of the main branch, vlab 5.1 build 3615, and snapicon works correctly. It links Qt 5.15.18 so it could be that your Qt version (5.15.19) has some changes that cause the issue (but I'm not sure). Alternatively, the latest version of the main branch has vlab daemon running always even if browser quits. You could try running pkill vlabd before launching browser to ensure vlab is not calling an older version of snapicon. Otherwise, try switching to the migrate/qt6 branch :) It seems to work better with the latest macOS