search

melowntech / workshop

Workshop of Melown 3D stack
10 stars 3 forks source link

Segmentation fault (core dumped) #31

Open Tekn1ck opened 4 years ago

Tekn1ck commented 4 years ago

Hi,

We have managed to take our OBJ files and process them into VEF format with all the LODs and manifest files. We have also managed to successfully convert from VEF to VTS tiles and create tilesets for small areas.

Below is a smaller dataset of 36 tiles up and running. image

image.png

Our problem is that when I scale up and put a lot of tiles in I am getting random Core dumps (Segmentation fault (core dumped)) and I have no idea how to troubleshoot. If it failed on the same tile every time I could go back and reprocess that one and it would be fine. All websearches point to the Segmentation faults being connected to memory and write permissions. We are running this on a VM online rather than a physical machine. it is a 32 core 96gb instance.

I don't think it is the 3d models or the manifest - but perhaps I have overlooked something.

Do let me know if there is anything you think we may have missed and I would be very grateful for your assistance - I feel like we are very close indeed to getting this up and running!

Core dump1.txt Core dump2.txt Core dump 3.txt core dump 4.txt manifestjson.txt renamed to .txt instead of .json because json uploads are not supported on github

I have also sent this report to info@melown.com if you would like a more in depth conversation.

EDIT: SOLVED!

Turned out that it was an obj file missing. We found out which one by running on a single core instance and it dumped at the same tile every time. It would be super to have a better set of error messages when things fail ;)

ladislavhorky commented 4 years ago

Hi Tekn1ck,

indeed, running single-thread with --log.mask ND or --log.mask ALL would be my first advise. It would run for ages but might give more insight.

As you are legitimately the first (congrats !) to create or report on their own VEF, there are not yet many (or any) checks in the code as the data has been so far always consistent ;) . We will see if with @vaclavblazek we can add some basic ones like you described.

Thanks for the report! Ladislav

Tekn1ck commented 4 years ago

thanks - we were very pleased with our progress. We are also going to build in our own data integrity checks too ;)

vaclavblazek commented 4 years ago

Since there's a SEGFAULT it is an actual error. There's probably missing a validity check. I'll look into it.