zed backup to stdout or existing file

[FrankFoerster24]

Hi,

I would like to send the backup to some subsequent processing, eg. compression, encryption, and other voodoo before hitting the final backup target. I would very much like if I could make it write to a previously created named pipe which I could read from with other programs. Maybe a --force overriding the check would already do the trick.

Alternatively, sending the output to stdout to pipe it to other tools could also work.

Kind regards,
Frank.

[bison]

Thanks for filing this! The v0.13.0 release should support this now.

[FrankFoerster24]

Hi,

thank you very much for the swift implementation. However, for completeness I'd like to add that I still have the use case to send the backup data to an existing file (which is AFAICS not possible right now).

This is my use case: I'd like to run backups using the provided docker images which don't have much other utilities, not even a shell to > redirect the stream on stdout AFAIK. I'm not keen to run my own images. The pattern I usually would use is that I send the data to a named pipe (created by mkfifo on a shared volume in an init container), have the zed container write to it and another container in the same pod read from it for further processing (compression, encryption, etc.). With zed refusing to write to the existing named pipe I cannot do that.

I'll find workarounds, but the described pattern would be fit most perfectly for me. Let me know how to proceed (reopen this issue, create a new one, go away and leave you alone... ;-)).

Thanks again and kind regards,
Frank.

[miparnisari]

It seems that we broke support for this in v0.28.0. Reopening.

[12:21:05] ~/Documents/GitHub/zed (main) $ zed backup create -

// ommitted for brevity

12:21PM ERR terminated with errors error="failed to store cursor in lockfile: cannot create temp file: open /dev/stdout.lock2654921088: operation not permitted\nsync /dev/stdout: inappropriate ioctl for device"

@ivanauth i left a comment to that effect in #688 (comment)

[ivanauth]

Fix is in #691 - drops the lockfile/Sync/Close path when the destination is -. Adds a TestFileEncoderStreamingToStdout covering the regression.

PR #690 separately adds tests for printBackupFileLocation per the review on #688.