rhinstaller / isomd5sum Goto Github PK

https://buildd.debian.org/status/package.php?p=isomd5sum

The only failing architecture I had ready access to is powerpc, and can reproduce. testpyisomd5sum.py:

Implanting -> PASS
Implanting again w/o forcing -> PASS
Implanting again forcing -> PASS
Checking -> FAIL
Run with callback, prints offset and total
    0 - -131941381867520
    13496320 - -131941381867520
FAIL
Run with callback and abort after offset of 500000
    0 - -131941381867520
    13496320 - -131941381867520
FAIL

checkisomd5:

testiso.iso:   28d3308dc6c8e592f0587849007b738b
Fragment sums: 3d4d8dfb3959eae63e8f3c26eb464ae7edec1eab97796331d22c718ec9ed
Fragment count: 20
Supported ISO: yes
Press [Esc] to abort check.
Checking: -00.0%

The media check is complete, the result is: FAIL.

It is not recommended to use this media.

ISOs implanted on powerpc and checked elsewhere work fine, while ISOs implanted elsewhere and checked on powerpc fail, so it's definitely a check issue, not an implant issue.

Not sure what this could be. All failing architectures are 32-bit userlands, but other 32-bit architectures work (i386, armhf, etc).

i686 log:
https://koji.fedoraproject.org/koji/taskinfo?taskID=19912455

arm log:
https://kojipkgs.fedoraproject.org//work/tasks/2480/19912480/build.log

So size_t on some arches isn't big enough.

Hello,
Same Here!
For reproduce this, you can follow these steps:
Clone (not download) this repository, then run: make and make test, You see test is passed and It's Okay.
Then remove .git directory (to reduce amount of files) and run make test again, The test is Fail!

Also when I use checkisomd5 command with verbose option, I saw ;FRAGMENT that appends to the Fragment sums of the iso, In my tests, It only appends when amount of file is low!

Originally posted by @hosni in #18 (comment)

Back in the days when this was moved out of anaconda @katzj simply copied the files out of anaconda into this repository. During that process valuable commit history got lost which even though it would require an evil force push should be added back in (in my humbly opinion).
It would be great if someone with commit access could do this by replacing the current master
with squimrel/isomd5sum/tree/merged.
The one who does this also has to check if I included any evil changes that shouldn't be there.
The "merged" branch is basically a continuation of the isomd5sum anaconda history by the current master branch excluding 0bd08db because it was already implemented in 9404540 which was lost when the project moved and also excluding the "initial import" 07cff92.

In any case I'd love to hear your opinion on this. Thanks.

It can be seen on Fedora 27 when running the test:
https://github.com/rhinstaller/isomd5sum/blob/master/testpyisomd5sum.py
If I add 500K file to the working directory the test result is pass.

It seems that on 32 bit platforms implantisomd5, and also checkisomd5 are broken:
https://bugzilla.redhat.com/show_bug.cgi?id=1562146