https://bugs.kde.org/show_bug.cgi?id=522307

--- Comment #4 from [email protected] ---
Thanks for looking into this!

1. kwalletd6 is stable — confirmed via `ps -o pid,etime,cmd`, the process
   has been running continuously (4+ minutes uptime at time of capture,
   matching session length) with no crashes/restarts. journalctl shows
   zero crash/core/segfault entries for kwalletd6.

2-3. I captured a full dbus-monitor trace (including replies, plus
   org.kde.KWallet and direct traffic to/from kwalletd6's two D-Bus
   connections). It's quite large (~5M lines) — happy to extract and
   attach a relevant excerpt if you can point me at what to grep for.
   For context: the same capture contains 17253 ReadAlias calls and
   17254 CreateItem calls — almost exactly 1:1, suggesting kwalletd6
   is calling ReadAlias and CreateItem together as a pair, repeatedly,
   rather than these being two unrelated polling loops.

4. I took 5 gdb backtrace samples of the live kwalletd6 process while
   the CPU loop was active (2s apart). I don't have app-level debug
   symbols resolved yet (working on that), but interestingly, 3 out of
   5 samples show the main thread NOT in defaultCollection()/ReadAlias,
   but instead in:

     secret_item_create_sync ()
     secret_service_create_item_dbus_path_sync ()

   i.e. libsecret's CreateItem path, not ReadAlias. This might mean
   kwalletd6 is repeatedly creating a new secret item (possibly related
   to the kdewallet_attributes.json growth I noticed earlier in this
   investigation, which grows by one entry-cycle on each write), rather
   than just polling ReadAlias in isolation — or both might be part of
   the same retry cycle. I'm attaching the raw backtrace samples as-is;
   will follow up with resolved symbols (debuginfod or debuginfo
   packages) if that would help narrow it down further.

Full backtrace samples attached.

-- 
You are receiving this mail because:
You are watching all bug changes.

Reply via email to