Hi again Sergey, I have run this yocto test and I have even reverted my changes to the original one for this opkg-build script https://git.yoctoproject.org/opkg-utils/tree/opkg-build and just set the line https://git.yoctoproject.org/opkg-utils/tree/opkg-build#n170 to something like this: tarformat="--format=posix --pax-option=exthdr.name=%d/PaxHeaders/%f,atime:=0,ctime:=0" but I'm still getting these differences between mtime values.
Additionaly both archives can't be in different formats because diffoscope would show us this difference in headers. BTW. you can download these archives which are compared and check by yourself from here: https://autobuilder.yocto.io/pub/repro-fail/oe-reproducible-20230724-5av30jl8/packages/ There are two directories reproducibleA and reproducibleB which stores these ipk pakcages. I have to admit that right know I have no clue what can cause this to happen and only for these 19 packages... Regards, Piotr ________________________________ Od: Sergey Poznyakoff <g...@gnu.org.ua> Wysłane: wtorek, 25 lipca 2023 14:13 Do: Piotr Łobacz <p.lob...@welotec.com> DW: Paul Eggert <egg...@cs.ucla.edu>; bug-tar@gnu.org <bug-tar@gnu.org> Temat: Re: posix atime/ctime changes despite mtime being set > as you can see on this output the only > differences are between timestamps e.g.: > > > 7 -rw-r--r--···0········0······· 7 -rw-r--r--···0········0······· > ·0·····7339·2012-02-16·20:24:4 ·0·····7339·2012-02-16·20:24:4 > 2.000000·./usr/src/debug/acpid 2.515701·./usr/src/debug/acpid > /2.0.34-r0/acpi_ids.c /2.0.34-r0/acpi_ids.c Then I'd suppose that first archive (the one shown on the left size) was created in ustar, gnu or similar format - the one that stores timestamps with a seconds precision. In contrast, second archive (the one shown on the right side) is in POSIX.1-2001 format, which stores timestamps with a subsecond precision. Regards, Sergey
