gtk: Defer browse popover teardown and fill - #18
Conversation
Windows still crashes in gtk-4 with c0000005 after the earlier sidebar reconcile fix. Report.log faults at a new gtk offset while Browse-all is in play. Do not popdown/unparent the browse popover from inside row-activated, and do not build the full ListBox before the popover is mapped. Also keep the Browse-all action off the sessions ListBox signal stack, and lower the visible row cap.
Fork PR CI fails Generate agent bundle with exit 65 on macos/ios. Disable codesign for the AgentBundle xcodebuild step (same as the main Luma build) and add a shared AgentBundle scheme so clean checkouts can resolve -scheme AgentBundle.
|
can't seem to fix the failed checks (im new to all of this) |
|
Thanks! Did you manage to build and run this locally? The .ps1 scripts should hopefully be helpful. Should only need MSVS (Community Edition will suffice) plus vcpkg and the Swift toolchain. |
@oleavr I use VS Code but my Swift toolchain is messed up I was not able to manage to build and run locally. Would I MSVS? |
|
@oleavr Hi just to make you aware I am having fatal issues with my laptop that I use with the coding workflow and seem to not be able to fix. Not sure when I will be able to resume. If you want, I can close this PR, and then if I fix my laptop, I can re-open or open a new PR. as of right now I can't do anything like compile or run the agentic agent. |
Apple CI fails Generate agent bundle because PackageInstallOptions and BuildOptions were removed from frida-swift. Pass the same settings as direct parameters on install()/build() instead.
Apple CI was cloning tip-of-main frida-swift (new Generated API) while Linux/Windows used Package.resolved (old options: API). The API migration commit fixed one side and broke the other. Restore the known-good PackageInstallOptions/BuildOptions call sites and pin frida-swift to 467d0360 everywhere, including the Apple CI frida-swift checkout used with the local FridaCore.xcframework.
Windows application still crashes in gtk-4 with c0000005 after the earlier sidebar reconcile fix. Report.log faults at a new gtk offset while Browse-all is in play.
Do not popdown/unparent the browse popover from inside row-activated, and do not build the full ListBox before the popover is mapped. Also keep the Browse-all action off the sessions ListBox signal stack, and lower the visible row cap.
@oleavr still crashing please review changes to make sure no other issues were introduced