arithmetic overflow for latest AADR on systems with 32 bit long

[libbro]

I am trying to use convertf on the latest version of the AADR (v62.0_1240k_public) and have been running into some issues. Whenever I run convertf -p parconv where parconv is:

genotypename:    v62.0_1240k_public.geno
snpname:         v62.0_1240k_public.snp
indivname:       v62.0_1240k_public.ind
outputformat:    ANCESTRYMAP
genotypeoutname: test.ancestrymapgeno
snpoutname:      test.snp
indivoutname:    test.ind

I run into the following error message:

parameter file: parconv
genotypename: v62.0_1240k_public.geno
snpname: v62.0_1240k_public.snp
indivname: v62.0_1240k_public.ind
outputformat: ANCESTRYMAP
genotypeoutname: test.ancestrymapgeno
snpoutname: test.snp
indivoutname: test.ind
read failure (length mismatch): Success
fatalx:
(inpack) bad data read (length mismatch) 2147479552 1140154008
Aborted

This is rather strange as I am using a file from a source that should be all good (literally from the Reich Lab). Any ideas what may be causing this issue?

[MatthewMah]

Please check your input file:
sha256sum v62.0_1240k_public.geno
7cf89c85c0fe8d4a1d92c211bd0e86796670a18e5e85e5cd275bd83eb6308279 v62.0_1240k_public.geno

[libbro]

Yep I got

7cf89c85c0fe8d4a1d92c211bd0e86796670a18e5e85e5cd275bd83eb6308279

[MatthewMah]

What OS are you running? What kind of CPU?

[libbro]

Debian 12. I have an Intel i5-10400

[MatthewMah]

I think the problem is the definition of long is different between your machine and where Nick has developed. I guess that your long is 32 bits, while ours is 64 bits. It's not apparent from your OS and CPU that this should be the case, but this would cause arithmetic overflow when reading the AADR genotype file (> 2GB, 31 bits + 1 sign bit).

I'm not sure what the timeline will be to fix this.