(Nota Bene: Many newer like xz/zstd/etc compressors have critical format
stability issues as compared to gz/bz2, see e.g.
https://github.com/facebook/zstd/issues/4548)

--
With Valediction,
Kamila Szewczyk (https://iczelia.net)

On 8/24/26 4:26 PM, Kamila Szewczyk wrote:
> Dear all,
> 
> My objection to this would be that the different on-disk format is
> different from the regular, single threaded format, leading to poor
> reproducibility.  There is no way to implement this in a way that
> ensures that pristine checkout is the same across multiple machines.
> 
> For example, a simple experiment is to take head -c ... | base64 and try
> to compress it with xz -T1 and xz -T2.  The checksums are different.
> Consequently the output of -T0 (and thus the tarball) depends on the
> $(nproc) of the machine.
> 
> The question left to be answered is whether this matters; to judge this
> I have forwarded the e-mail to Karl.
> 
> --
> With Valediction,
> Kamila Szewczyk (https://iczelia.net)

Attachment: OpenPGP_0xC868F0B6DE38409D.asc
Description: OpenPGP public key

Attachment: OpenPGP_signature.asc
Description: OpenPGP digital signature

Reply via email to