AF crashes @ launching but can be recovered by saving to prev. LV ver.

[mirode]

I seem to encounter the followng problem with my AF project (>1500 VIs, ca. 25 actors, LV 2014):

After making major changes to a project, like moving methods to diffrent classes, changing multiple VIs, etc. and trying to run the project, LV repeatedly crashes, the likehood it it happenning being proportional to the complexity of changes applied to project.

Restaring the PC, moving project to diffrent location or mass compile do not help. Once it happens it keeps repeating every time. There is no broken arrow or anything, just crash upon launching, giving diffrent reasons each time.

The only thing that proved to work is to save the whole project under previous version of LV, reopen it and save back again. Given the project complexity as above, it takes a minimum of 1h, if there are no additional issues with dependencies, conflicts etc.

It happened to me on 2 diffrent PCs (one using 64, the other 32 bit LV version) several times in the past year. I remember having experienced this issue before with LV2013 a couple of times.

I wonder if any of you have experienced similiar behavior and if there is a better workoround or diagnosis possibility to prevent such a situation?

Thanks in advance.

[AristosQueue (NI)]

Please escalate questions like this through AEs and NI tech support. They'll be better able to help you untangle this. I don't have any low hanging fruit to offer you.

[CaseyLamers1]

I have seen a similar situation, but it hasn't happened to me in quite some time. I was deploying to a cRIO from the dev environment. It would claim a certain method was loaded with errors and then stop deploying. Opening the method showed that it had a broken run arrow. Deleting a wire and re-wiring it would fix it. A subsequent deploy would result in something else saying it loaded with errors.

My "work around" is that I build rtexe's. The obvious drawback is that you can't probe wires at runtime. The fix there is to write things to files. As such I have an error logger actor that I spin up at the beginning, give the enqueuer to my actors and then write to it in Handle Error.vi.

I don't think I ahve had a problem with builds breaking even after doing fairy major refactoring.

[cirrusio]

OK...I too have experienced this and have been frustrated by it.  Aaaaannnnnddd, I learned a cool little trick that addresses at the very least Casey's problem and I believe it should help you with yours.  When you start getting these broken run arrows when deploying from the development environment, clear the compiled object cache, close the project down and restart.  If you are still getting the broken run arrows or crashing, try deleting the compiled object cache (located in Labview Data under VIObjCache) and reboot LV again.  I don't fully understand why this works, but I have not had any of the heavy crashing that mirode is experiencing since I started to do this.

Hope this helps.  Cheers, Matt

[mirode]

mtat76, your post raised my immense hope that there is a better workouround but I just bitterly convinced myself that it does not help in my case at all.

After implementing a series of changes in the project I am not able start it now again (do not get any broken arrows but LV crashes several seconds after launching the starter VI, as before). The crash report mentioned some accessor VIs that I have not touched for months as possible reasons. After modifying those VIs' execution parameters to no-reentrant (more or less just to see any change) the crash still occurs but now not telling me any specific suspisious object.

Each time it is some access violation.

I am going to (well, have to) do my "magic trick" and I am pretty sure it will work.

I will also start a new support request on that, as AQ suggested.

Thank you.