late package version collection can cause confusion

[Brian Murray]

The following OOPS contains some misinformation:

https://errors.ubuntu.com/oops/3680877e-5046-11e7-89b7-fa163e54c21f

The Date field is from before the package version in the report was accepted into -proposed. This likely happened because the .crash file was created and data collection, including adding the Package key, was done after systemd had been upgraded. This is alluded to in the UnreportableReason of the OOPS.

UnreportableReason:
Неполадка произошла с программой /lib/systemd/systemd-resolved, в которую были внесены изменения с момента её аварийного завершения работы.

It says that systemd-resolved was modified since the error occurred.

As we can see in bug 1621396 this ended up confusing the developer and they were concerned about their fix not working. It seems like apport or whoopsie should do something better here.

Ideas include not putting a package version in the Package field if the binary has been modified, or not uploading the crash to the Error Tracker if the binary has changed. The latter would require some additional work because the UnreportableReason is translated.

[Dimitri John Ledkov (xnox)]

i'm not concerend about false positives arriving in the errors tracket, but it would be nice if they were marked as guessed version with like a question mark when there is UnreportableReason.

[Brian Murray (brian-murray)]

Daniel van Vugt recently commented on a bug report about the errors bucket show a couple of crashes with the new package version and I'm concerned that those might actually be ones where the version collection was done after the package had been updated.

[Brian Murray (brian-murray)]

When looking at bug 1822395 it occurred to me that this could also be an issue if an application were to crash during the release upgrade process and the data collection were done after the upgrade.