slashdot hack and pattern matching

[ThomasWaldmann]

Discussed in #8674

^{Originally posted by callegar February 22, 2025}
Hi, I wonder if the documentation could be clarified with respect to root specification, the slashdot hack and pattern matching.

• Can the slashdot hack be employed also in root specification in pattern files?
• When the slashdot hack is used, are patterns matched against the full path or the stripped path? I think the first. If so would be able to specify a prefix that is applied to all the following patterns to reduce verbosity.

[PhrozenByte]

I believe this was already fixed in #9647, assuming you consider it a docs issue since you've labeled it documentation.

[ThomasWaldmann]

Correct, thanks for the reminder.

So, guess this issue is about whether it should be that way in borg2 or whether it should do matching on the stripped path.

[PhrozenByte]

So, guess this issue is about whether it should be that way in borg2 or whether it should do matching on the stripped path.

My two cents on this: I think it's good as it is. It's more powerful. Assume backing up /media/first-device/./ and /media/second-device/./, which intentionally merges the contents of both source paths into the archive's root path - we would lose the ability to apply filters depending on the source. Right now, we can tell Borg to include a subdir only from one source but not the other; if we matched against stripped paths, we wouldn't be able to do that.

To address the inconvenience of having to repeat the full path all the time, we could introduce a new pattern marker (using F below) that defines a common prefix for all following filters. Something like this:

R /media/first-device/./
R /media/second-device/./
F /media/first-device
- exclude/just/from/first-device

But this would add quite a bit of complexity to solve a (IMHO fairly minor) inconvenience.

[ThomasWaldmann]

Good ideas. Well, guess then we can close this, nothing to do.