On 8/26/25 18:11, [email protected] wrote:
On 27 Aug 2025, at 5:55 AM, Colin Percival wrote:
- new-prefix_test: its list of files is much smaller than the source archive.
But when I compared the file list they match exactly until new-prefix_test
file ended.
As you see there is no ".part" in its name.
new-prefix_test 36 MB 15 MB
(unique data) 205 kB 37 kB
That's very strange. Is there anything interesting about the next file in the
old archive after the new archive stopped? I'm wondering if somehow tarsnap
failed in the middle of copying the archive... if that happened you should
have seen an error message though.
Internet disconnection definitely had happened. Also, the next file was
just a file like it the last one copied - another .git/objects folder file.
What I am wondering is why there was no .part in the name of the archive.
Ooh, I think you might have uncovered a bug here.
Not having a .part is because tarsnap "finished" the archive. But I think
there might be an edge case where *reading* the old archive failed but the
archive *write* path didn't fail because the network came back up.
I'll have to look into this.
--
Colin Percival
FreeBSD Release Engineering Lead & EC2 platform maintainer
Founder, Tarsnap | www.tarsnap.com | Online backups for the truly paranoid
