-
Notifications
You must be signed in to change notification settings - Fork 1k
Segmentation fault when opening a contact #6412
Comments
From the backlog:
It's a bit hard to say what pointer in I'm unable to reproduce this at either v1.17.3-316-g12fc33ee or current master tip (65c42eb), my steps are:
Does this seems consistent with your repo? |
I don't think any of those are the issue. My friends list was mostly static and this usually happened while my friend was offline.
If I follow your exact steps, the issue doesn't reproduce on my end either.
No, this shouldn't take 100 times. It happens on my end when I click on the same contact just two times. 100 times is overkill. And if I already opened the contact with "Open each chat in an individual window" enabled and then I disable that option, clicking on that contact again will make the process crash again |
I wasn't trying to imply that the settings are required to be used together, the crash is a bug and the dependency of the settings is reflected accurately in the menu already. Interesting on my host system (Arch) with I still couldn't reproduce the issue even with "Open each chat in an individual window", with ~100 attempts. Checking on Fedora 35 though, it happened first try on the latest commit. They're both using Qt 5.15.2, so it doesn't seem to just be a Qt bug that was fixed between versions. I'll keep investigating now that I have a repro. |
Thank you |
full backtrace here:
so nothing like us just referencing a null pointer. It's crashing inside of(?) QWidget::show, but the bt just points to the first line of the definition. The entire function is this:
Will analyze more tomorrow, but looks like the |
Brief Description
OS: Fedora 35
qTox version: qTox v1.17.3-316-g12fc33ee
Commit hash: c0e9a3b
toxcore: toxcore version 0.2.12
Qt: version: 5.15.2
…
Reproducible: Always
Steps to reproduce
Observed Behavior
Process crashes unexpectedly.
Expected Behavior
Regardless of how many times I close and re-open the exact same conversation history window with my contact, the qTox process should never crash.
Additional Info
Fedora doesn't have a package with the debug symbols for qTox. As such, I cannot provide a reliable backtrace with GDB, sadly.
For what it's worth, as unreliable as it is, the backtrace that I did get can be found at: https://pastebin.com/a5xL4az1
I don't have much technical expertise but this looks like a nullptr de-referencing issue to me.
The qtox.log can be found at https://pastebin.com/3FGYDCxJ
Please note that both the backtrace and the log are set to be auto-deleted after 1 calendar Year since the moment they were created (around the same time that this issue was created).
The text was updated successfully, but these errors were encountered: