Replacement results in "Invalid cross-device link" #145
Comments
|
The most similar issues to this one are:
|
|
Sorry @issuedigger, these don't seem to be related :-) |
|
Hi @lukasjuhrich , thanks for your report! Apologies for @issuedigger, it's an experiment and safe to ignore. To debug this further:
Initial thoughts:
My FS setup/OS is different from yours, but I've run srgn on a vanilla Ubuntu 24 machine and it worked fine. I'll try that again if you can help me with reproduction steps! |
|
Oh and to try and clear up any confusion more: you titled the issue "two scopes", but your command actually uses 3:
It then also has one action:
If this is not what you intended, do you feel UX could be better? Documentation? Some fail-safe errors or warnings which could've helped you? |
|
Thank you very much for your insightful response! Indeed, the error is with replacing. As you conjectured, I did not intend to replace anything. reproductionWe can simplify the example by creating a TRACE logs (not really revealling anything new)However, The
UXI'm not sure how to solve this, but let me present what I think the problems are. First, that I can "push forward" the search argument (in this case, Second, I think it is very dangerous to have a destructive operation so "close" to a nondestructive one. I can think of two scenarios by which one can switch from
I tend to be afraid of tools which punish me for mistakes like that, because I tend to make many. I feel this is contrary to the "fast" and "light-weight" tool that srgn aims to be (and for me already is) if I have to be careful like that all the time; this is nontrivial cognitive load. I can think of two options:
Both require changing the default to "don't replace unless told to", which might require a deprecation period. Re documentation, I think changing the semantics is more desirable than documenting the existing semantics better for the reasons I've outlined. what I actually wanted to achieveI guess this would bettor belong in a disussion instead of an issue, but since you asked, let me tell you of my use case: I wanted to find all the |
lukasjuhrich
changed the title
|
i'm having the same issue with a simple rename operation ( |
|
i believe the issue is due to the way the dependency Lines 541 to 543 in 9238fc1 we declare
probably because probably, we should be using the |
|
Thanks @lukasjuhrich and @igor47 for your detailed reports. I'll look into it more soon. The As for the UX, agree! At the very least, the two positional arguments of Re: how Re:
Yeah don't think that's possible! If at all, it'd require substantial regex use. The without part is not really supported, |
`persist` of `tempfile` does not work across device boundaries, so if the device backing the temporary directory is e.g. `tmpfs`, then `persist` fails. It actually had no useful benefit, so we can just remove it and write to the destination directly. Reproduced bug and confirmed fix on a Debian machine with: ```console $ sudo mkdir -p /path/to/temp_dir $ sudo mount -t tmpfs -o size=100m tmpfs /path/to/temp_dir $ sudo mount | rg 'temp' tmpfs on /path/to/temp_dir type tmpfs (rw,relatime,size=102400k,inode64) $ export TMPDIR=/path/to/temp_dir $ cargo run -- --python class a whatever ``` `TMPDIR` env var is respected by `tempfile`. The above `cargo run` fails without this patch. (There is an end-to-end test for file writing, but that has never failed for some reason... seems like none of my machines (including Ubuntu 24) or CI exhibited the issue). Closes #145
As soon as a temporary file is involved, the temporary file seems to cause problems:
stdout / stderr
FS setup
/tmp tmpfs tmpfs rw,nosuid,nodev,nr_inodes=1048576,inode64/home /dev/mapper/vghdd-lvhome ext4 rw,relatime,data=orderedHow can I debug this further?
The text was updated successfully, but these errors were encountered: