[Bug 2125535] Re: makeself archive failing md5sum validation
Dimitri John Ledkov
2125535 at bugs.launchpad.net
Fri Sep 26 21:17:29 UTC 2025
>From access user in phoronix article:
Seems like the dd options are the problem. If you use ibs instead of bs
at all places in the dd wrapper it works as expected with uutils too.
bs= should set both ibs and obs.
Something strange going on with coreutils too in this case as it reports
the wrong amount of bytes read.
Code:
❯ sh -c '(dd ibs=20487 skip=1 count=1; dd bs=4194304 count=1; dd bs=2441216 count=1) < VBoxLinuxAdditions.run | md5sum'
1+0 records in
40+1 records out
20487 bytes (20 kB, 20 KiB) copied, 6,6885e-05 s, 306 MB/s
1+0 records in
1+0 records out
4194304 bytes (4,2 MB, 4,0 MiB) copied, 0,0119912 s, 350 MB/s
0+1 records in
0+1 records out
2420729 bytes (2,4 MB, 2,3 MiB) copied, 0,00790668 s, 306 MB/s
201ccc47587bb1c28745279e8b7fdd30 -
The last one should be 2441216 bytes. Changing the first skip to instead writing to /dev/null produces the correct value
Code:
❯ sh -c '(dd ibs=20487 of=/dev/null count=1; dd bs=4194304 count=1; dd bs=2441216 count=1) < VBoxLinuxAdditions.run | md5sum'
1+0 records in
40+1 records out
20487 bytes (20 kB, 20 KiB) copied, 8,5465e-05 s, 240 MB/s
1+0 records in
1+0 records out
4194304 bytes (4,2 MB, 4,0 MiB) copied, 0,011503 s, 365 MB/s
1+0 records in
1+0 records out
2441216 bytes (2,4 MB, 2,3 MiB) copied, 0,00491868 s, 496 MB/s
201ccc47587bb1c28745279e8b7fdd30 -
And finally uutils with ibs instead of bs for the last block
Code:
❯ sh -c '(./coreutils dd ibs=20487 skip=1 count=1; ./coreutils dd bs=4194304 count=1; ./coreutils dd ibs=2441216 count=1) < VBoxLinuxAdditions.run | ./coreutils md5sum'
1+0 records in
40+1 records out
20487 bytes (20 kB, 20 KiB) copied, 0.00441375 s, 5.1 MB/s
1+0 records in
1+0 records out
4194304 bytes (4.2 MB, 4.0 MiB) copied, 0.0127932 s, 350 MB/s
0+1 records in
4727+1 records out
2420729 bytes (2.4 MB, 2.3 MiB) copied, 0.0062951 s, 403 MB/s
201ccc47587bb1c28745279e8b7fdd30 -
Last edited by access; 26 September 2025, 01:25 PM.
--
You received this bug notification because you are a member of Ubuntu
Foundations Bugs, which is subscribed to coreutils-from in Ubuntu.
https://bugs.launchpad.net/bugs/2125535
Title:
makeself archive failing md5sum validation
Status in Release Notes for Ubuntu:
New
Status in coreutils-from package in Ubuntu:
New
Status in makeself package in Ubuntu:
Confirmed
Status in rust-coreutils package in Ubuntu:
New
Bug description:
Still trying to find causes, but this sympton:
for a 25.10 guest image, the virtualbox guest additions install
(VBoxLinuxAdditions.run from the official ISO) fails checksum. The
Linux installer script is a makeself self-extracting archive (makeself
is a Ubuntu package, though I don't know if the VirtualBox team uses
Ubuntu to generate the archive). The output of trying to run this
script is:
# ./VBoxLinuxAdditions.run --noexec
Verifying archive integritiy... 100% Error in MD5 checksums: c8ea742682d7024e581f916c763818c9 is different from 201ccc47587bb1c28745279e8b7fdd30
The 201ccc signature is the one embedded in the script. On a 25.04
system, there is no checksum complaint:
# ./VBoxLinuxAdditions.run --noexec
Verifying archive integrity... 100% MD5 checksums are OK. All good.
Uncompressing VirtualBox 7.2.2 Guest Additions for Linux 100%
I was sure this was due to the switchover of the waste-of-time
coreutils rust rewrite (please ignore the editorial comment!)... but
copying over coreutuils md5sum from 25.04 does not eliminate the
error.
Tar on both systems is 1.35.
md5sum on today's (23 Sep) 25.10 reports: md5sum (uutils coreutils) 0.1.0
on 25.04 where there's no failure it's: md5sum (GNU coreutils) 9.5
but as noted, that exact binary moved to 25.10 to replace /usr/bin/md5sum does not improve matters there.
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu-release-notes/+bug/2125535/+subscriptions
More information about the foundations-bugs
mailing list