[Tracker] Triggers debugger assert on quit with debug builds

[x-ist]

Tested on hrevr1alpha4-44586. Works reliably. (will add more info later today)

[diver]

Can't reproduce in hrev44620.

[x-ist]

DebuggerStacktrace

[anevilyak]

That assert should only be getting tripped on a debug build of libtracker. Are you saying you see the same behavior with a release build?

[x-ist]

I tried debug build only, so I guess yes. :)

[anevilyak]

Replying to x-ist:

I tried debug build only, so I guess yes. :)

"Release" build generally implies without debugging. In any case in future it'd be helpful if you could be a bit more explicit about the fact that you're building with custom changes, as the ticket sounds as if you're encountering this on a plain nightly image, which isn't the case.

[x-ist]

Replying to anevilyak:

"Release" build generally implies without debugging. In any case in future it'd be helpful if you could be a bit more explicit about the fact that you're building with custom changes, as the ticket sounds as if you're encountering this on a plain nightly image, which isn't the case.

Sorry for that! I made a debug build of libtracker.so and Tracker. I thought it's clear from the screenshot. Otherwise Debugger couldn't refer the source code.

[anevilyak]

Replying to x-ist:

Sorry for that! I made a debug build of libtracker.so and Tracker. I thought it's clear from the screenshot. Otherwise Debugger couldn't refer the source code.

It was obvious that the screenshot was taken with a debug build, yes. That didn't in any way clarify that the crash occurred exclusively with a debug build, i.e. that you didn't actually try if it crashed with a regular one, which is what the ticket description initially sounded like.

[x-ist]

You're right! However my opinion was that if there's something wrong in the debug build, it concerns the release build as well - well could be I'm wrong.

[anevilyak]

Replying to x-ist:

You're right! However my opinion was that if there's something wrong in the debug build, it concerns the release build as well - well could be I'm wrong.

ASSERT()s compile to a no-op in a release build, therefore they won't be hit there. Whether they're actually correctly pointing out a problem or not isn't quite as cut and dried, particularly in Tracker where quite a number are left over from back in Be's day and haven't really been reevaluated to see if they're still correct after all the rewrites that have happened over the years.

TL;DR: That's not a good assumption to make.

[axeld]

In any case, if it is only hit on quit, and there is no visible problem in the release build, it's not an urgent problem. It's something that should be looked into in any case, though.