On 26/05/2024 22:08, Sylvestre Ledru wrote:
Hello,
Playing with the one of the *sum command, i noticed a weird behavior.
Depending on the order in the checksum file, we will get completely
different results.
For example, this code reproduces the issue:
----
touch a
touch "*c"
echo "First test case:"
echo "d41d8cd98f00b204e9800998ecf8427e *c" > in.md5
echo "d41d8cd98f00b204e9800998ecf8427e a" >> in.md5
/usr/bin/md5sum --check in.md5
echo ""
echo "Second test case:"
echo "d41d8cd98f00b204e9800998ecf8427e a" > in.md5
echo "d41d8cd98f00b204e9800998ecf8427e *c" >> in.md5
/usr/bin/md5sum --check in.md5
----
Will return:
----
First test case:
/usr/bin/md5sum: c: No such file or directory
c: FAILED open or read
/usr/bin/md5sum: WARNING: 1 line is improperly formatted
/usr/bin/md5sum: WARNING: 1 listed file could not be read
Second test case:
a: OK
*c: OK
----
with:
md5sum (GNU coreutils) 9.4
Is that a bug?
I think I understand the technical reason even if I didn't check the
code. I guess the parser will look at the first line
and think that it the "double space format".
It's not a bug, it's an unsupported edge case for auto handling BSD reversed
format.
The edge case considerations are documented in:
https://github.com/coreutils/coreutils/commit/7da844cc8
To state explicitly, it's only when the first file name presented to --check
in a BSD reversed format checksum file, has a ' ' or '*' first character.
That's quite the edge case, which should not hit in real world use cases.
cheers,
Pádraig
