Michael Eischer

I have to say I _seriously_ dislike rolling our readdirnames implementation. There's a reason such things are part of the standard library. As inode numbers are expected to be unique...

> But it's not documented what should happen if it returns real struct stat with zero st_ino. At least the linux kernel header files say the following https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/include/linux/fs.h?h=v6.17.8#n3365 : >...

> One better way to do this might be calling `syscall.ReadDirent` with a really large buffer size, and then using `syscall.ParseDirent` on the result. That will then also skip directory...

Thanks a lot for the investigation into this bug! > This is what I meant: [master...vasi:restic:readdir2](https://github.com/restic/restic/compare/master...vasi:restic:readdir2) > > It does work, but I'm really not sure if I prefer it!...

> Let's give them a couple weeks and revisit? Sounds good.

> When using the `snapshots` command it would be useful if one could filter by date (before or after a certain date, or within a range). > > Other commands...

See https://github.com/restic/restic/issues/5193 for a related option for the `forget` command.

> True, but I don't think that the usual case will be lots of parent snapshots. The standard case will be exactly one parent and backup runs with more -...

> * The size of the corrupt pack file was wrong in the index The index contained the correct pack size, the pack file itself has a wrong size. >...

> Do you have a sample of the log output? Then we can improve the retry handling to quickly skip the missing parts of the file. I've just checked, our...