I have been playing with the software all evening with an empty file I wanted to try some of the other things that have changed, so I kept it running to see what would happen. If I did not create the file, kstars would not start. Should the file be created on start-up automatically?
This makes a little more sense now. I did play with 1 setting in the default profile 2 nights ago when I was experimenting with settings to learn more. I did not set it back before ending the session. Might there be a reference stored somewhere in the system now that is looking for that modification, and the ini file did not get written? Now on start-up, it is looking for it, and it cannot find it.
Some speculation there, but it should be reproducible then by starting with all the defaults and no .ini file. Make one change to the default profile. Exit kstars, and see 1) if the ini is created, and 2) if it is not, do you get the crash?
I actually had a good night last night of using some of the new features, and managed to capture 6 different objects. Today, I pulled master again and rebuilt. I removed the empty SavedHFRProfiles.ini file, and kstars has been running normally. I picked up a new Ryzen a few weeks back and can now build with -j15. It builds kstars in about 2 minutes 30 seconds. Fun, and much easier to try new software quickly.
Wait, so whatever strange thing was wrong, it went away on its own???
Both of the crash reports you posted were centered on the same issue and neither one crashed on the loading of the saved files. Both crashes occurred when setting up the convolution filter. I still don't understand why the QVector that was set up correctly for the convolution filter initially caused a crash when it was told to clear as your one error report stated, or that it crashed when allocating the QVector when it was being set up the second time. But you say the error just went away?
I would like to be able to find out what happened. . . it could be some sort of QT bug.
There are always variables. In this case a code update (git diff 7ab6c95a63ead12677997e028521c23b4ee724cc) which has plenty of Qt changes, though none directly related to the handling of this file. There were also system updates on my OS, so there could have been buggy Qt5 v5.15 packages that were fixed as well.
It would be nice to know what made the difference, but if nobody else was having problems, I may have just had the wrong combination of transient software.
I am going to try bisecting the merged code bits to see if a build crashes on any of them any more.
I went back 3 days worth of git hashes, compiling and testing each one. I was unable to reproduce the problem of kstars crashing at startup when the file did not exist. I think we need to chalk this one up to system packages in that case, and move on. Hy and Rob, thank you both for looking at this.