Quote:
[...] on the source side, not on the target side. In other words, tar obtains the inode times, reads the file into the accumulation of the output file, then resets the inode times to the original.
Yes drl you are correct. But if an archiving utility like tar includes the ability to carefully restore the original's atime (which requires extra work, i.e. modifying the inode after the file is read), it would be bizarre if the programmers didn't also include the ability to set the extracted copy's atime (which requires absolutely no extra work, because tar always sets the extracted copy's time correctly, and assuming it also sets the atime to something, it is no extra work to set it to the correct value).
Quote:
Some filesystems are set not to even record the access times, since it is often of limited value and causes disk activity.
Correct. And even if I tar files from such a filesystem (which hence always have an atime long in the past), the atime of the extracted copy is the current time. And also, even if I tar files and *extract* them onto a filesystem that does not record the atime (mount option noatime), the atime is set to the extraction time.
Quote:
If my supposition is correct, then one could use touch to reset the access times on the target -- tedious, but doable.
But how? Presumably they are stored in the tar archive, but where? And if I wanted a tool that reads stuff from a tar archive, the tool "tar" would be my choice :-)
We are talking about 10000 to 100000 files here. Possibly there are scripts that store and later reset the atimes, but I don't know of any.