Crash tracking backwards

PeTrack crashes while tracking (performing no recognition). This sporadically happens e.g. when tracking backwards with project ias-7_ped:/experiments/2020.06.16_Juelich_IEK6H3_Engstelle/export/simplylife/20200616_09561114_100_l020_g1_h-_400_cam8.pet

assigned to @d.kilic

By Boltes, Maik on 2020-10-07T11:30:08 (imported from GitLab project)

It worked flawlessly when I started tracking & reco through calculate all on my local machine:

@m.boltes Is it important that I don't do Recognition?

By Kilic, Deniz on 2020-10-08T08:29:05 (imported from GitLab project)

I tried it without recognition to narrow down the origin. Unfortunately it happens sporadically.

By Boltes, Maik on 2020-10-08T16:32:40 (imported from GitLab project)

Seems to be a stackoverflow. I haven't investigated it yet, but this might need a change of the control flow of PeTrack, especially the forward->updateImage->done "Loop", which is continously adding to the stack during playback. (The exception was thrown in player::updateImage)

But I'm not sure if this is everything, since this time it crashed when tracking backwards after about 8000 Frames processed. It should have done so when calculating all as well.

By Kilic, Deniz on 2020-10-13T10:33:09 (imported from GitLab project)

I got a stackoverflow just by playing the japan marker video forwards and backwards, this time in forward (part of the forward->updateImage->done "Loop" as well) So I am quite certain that this is a problem which needs fixing, even though I can't be sure it is the one which occured originally.

It doesn't happen with trackAll aka. calculate all, because that just calls mPlayerWidget->frameForward() in a loop.

(I also looked at the callstack this time and it is the mentioned "loop" cluttering it to a degree, where I can't even see the whole one)

created branch 72-crash-tracking-backwards to address this issue

added Doing label

mentioned in merge request !59 (merged)

closed with merge request !59 (merged)

mentioned in commit ed2bfb06