Makes tar-fs able to pack files that are actively being written to. #88
Conversation
|
Can you explain a bit more what this fixes? |
| @@ -138,7 +138,7 @@ exports.pack = function (cwd, opts) { | |||
| var entry = pack.entry(header, onnextentry) | |||
| if (!entry) return | |||
|
|
|||
| var rs = mapStream(xfs.createReadStream(path.join(cwd, filename)), header) | |||
| var rs = mapStream(xfs.createReadStream(path.join(cwd, filename), { start: 0, end: header.size > 0 ? header.size - 1 : header.size }), header) | |||
There was a problem hiding this comment.
Ah, I remember what this does now (it's been a while).
tar.pack('some_dir') works by getting a file from some_dir, creating a ReadStream for that file, then piping the ReadStream to a tar entry, which triggers this whole process for the next file in some_dir upon completion of the pipe, until all files from some_dir are packed.
However, if a file in some_dir is continuously being written to (e.g. a log file), that file has no EOF, and piping it's ReadStream to an entry will never complete, holding up the whole packing process and causing the tar file for some_dir to never be made.
This change adds start and end parameters to the ReadStream of a file based on the files size at that time. Adding these parameters ensures that piping the ReadStream to an entry will reach completion even if the file has no EOF, allowing the tar packing process to continue.
|
Thanks for explaining this. If you can remove the lock file then this is good to go! :) |
Whoops! 😅
|
Went ahead and removed the package-lock.json. I think that got in by accident 😅 |
|
2.1.0 |
|
Thanks for the contribution! Appreciate it. |
Pretty simple change that allows tar-fs to be robust in more conditions, while maintaining its expected behavior.