FET sources package size

[Liviu Lalescu]

I have an enigma: somewhere between FET-7.8.2 and FET-7.8.3, the size of the tar.xz package increased  from 11.9 MiB to 12.8 MiB. I checked and the difference appeared on 27 March 2026, out of nowhere, because the size of the files is nearly the same.

Could it be because of an update of the operating system programs?

[Volker Dirr]

Maybe extract FET-7.8.2 and compress it again. If you think the update of the OS is the reason, then then new compressed FET-7.8.2 should be larger too.
I will try and search myself now.

[Volker Dirr]

I guess the reason are the many new example and test files.

[Liviu Lalescu]

I put on a temporary link: https://lalescu.org/fet/ (I'll remove it soon), the position where this difference appeared. No new files, only slightly modified ones.

I tried to pack again the small file, it was even smaller.

[Volker Dirr]

Maybe they use in their algorithm somewhere an sorting function or hash that is not always exactly the same? I remember that we had "similar" problems some time ago in FET, so that it produced different timetables even with the same seed.

[Volker Dirr]

I read that you can avoid that problem by using the "--reproducible" flag.

[Liviu Lalescu]

Please unpack my two files above and pack them with "tar -cvvJf archive-name.tar.xz directory-name" on command-line and let me know.

[Volker Dirr]

i kdiffed. there are small differences in the archives.

the file sizes of the new compress files (compressed by me) are:
11.863.740 for the 20_22 file
11.859.196 for the 12_40 file

[Volker Dirr]

so my files are much smaller than yours. I use tar (GNU tar) 1.35

[Volker Dirr]

The algorithms i am using are:
xz (XZ Utils) 5.4.5
liblzma 5.4.5

[Liviu Lalescu]

This is really interesting!

I use the same tar version, but I think the thing that matters is xz:

Code Select Expand

goghi@localhost:~> xz --version
xz (XZ Utils) 5.8.2
liblzma 5.8.2