CloudCompare forum

Hey Daniel,
i've stumbled again across the bug of not being able to save/ open large datasets in 2.8., where I get the error message: "Read error (corrupted file or no access right?)".
The dataset is approx. 2,5 billion points, split into 12 clouds. Furthermore CC crashes when I try to merge the single clouds.

Cheers,
Arie.

System specs:
Win 7 Pro 64bit
Xeon E-5 1650 V2
128 GB RAM
Nvidia Quadro M4000

Wow, 2.5 billion?!

I never tested CC with such a huge cloud, so it doesn't seems that strange that it crashes ;). At least for the merge method it's quite logical that it crashes (I think the theoretical limit for a single cloud is 2 billion points anyway). We should at least add some warning messages in this case...

Nevertheless the file issue is a lot more strange (especially since there are 12 clouds, so I think each cloud is not that big?). What is the generated file size? (in bytes)

Unfortunately I deleted the "corrupted" file with all clouds loaded.

But I tried splitting the file into two single ones, each containing six pointclouds (where each cloud has between approx. 100 - 500 million points).
Again both bin files can't be opened.

File 1: 21.394.142.519 Bytes
File 2: 17.737.391.296 Bytes

Cheers.

If it's not too long, is it possible for you to repeat the process of saving 1, 2, 3, up to 6 (potentially) clouds and determine after which file size CC fails to load the file?

There was a 4 Gb limit at a time with older versions of Windows, but it doesn't exist anymore (I can safely read 6 Gb files on my side).

Weird things happening here...

Following your suggestion, I imported the single clouds incrementally, each time saving it as a new file.
Now I can open the .bin file with six clouds loaded (File 1). But, if I import all the six pointclouds at once and save a bin file, it becomes corrupted.
What's going on?!

That is indeed weird!

Is the final file size exactly the same? And are the clouds organized the same way in the DB tree?

Well, after importing all files at once, I place them all in the first folder, mark the folder and save as bin.
Importing the files seperately, I also just place the newly imported single cloud in the same folders as the previously imported clouds.

Regarding the filesize: I have to check on that, because I deleted the corrupted file. I'll get back to you on that in a moment.

Filesize is identical.

Even more puzzling... It would be interesting to do a byte-to-byte comparison, but with files so huge it might get dirty. It's really strange.

If you want, I can bring the dataset with me to VirtArch2017...