Copy stdin to stdout and stderr, unbuffered

A quick tool to debug a pipe. And more.

Please note: Version 1.x differs from 0.x that -q no more defaults to LF if no other line modifying option is present.

This makes it easier to use in case you want to use -c without other options.

If -v, -l or -u is present, the default is LF again.

Usage

git clone https://github.com/hilbix/unbuffered.git
cd unbuffered
git submodule update --init
make

Examples

Record a pipe to a file. Uses option -q to remove LFs from incomplete lines:

producer | unbuffered 2>>file | consumer

Record a pipe to a file with timestamping, incomplete lines are marked with +:

producer | unbuffered -uv 2>>file | consumer

Debug some pipe step: Record stdin and stderr into debug.log:

input to producer | unbuffered -uvd producer 2>>debug.log | consumer

Hexdump stdin/stdout of a command to stderr

unbuffered -ki0 -x -p'Ia ' -r'Ib ' unbuffered -x -p'Oa ' -r'Ob ' command args..

Hexdump stdin/stdout/stderr of a command to FD3 (which can be redirected to a file like here in the example: /tmp/dump)
stdin, stdout and stderr are passed through to/from the command unharmed
note that the sequence of the unbuffered commands is important here

3>/tmp/dump unbuffered -x -i2 -w2 -o3 -p'Ea ' -r'Eb ' unbuffered -x -o3 -p'Oa ' -r'Ob ' unbuffered -x -ki0 -o3 -p'Ia ' -r 'Ib ' -- command args..

Option -h prints a complete list of options, like:

option -b: buffer incomplete lines. Give twice to buffer all fragments.
option -c: cat mode, do not duplicate input, instead output stderr to stdout.
option '-n': output line numbers
option -x: hexdump output

Notes

To see the difference between no buffering, -b and -bb use following example (this are 3 lines!):

show() { echo "unbuffered $*:"; { echo -n "partial line :"; sleep 1; echo -n ": complete
partial :"; sleep 1; echo -n ": multiline :"; sleep 1; echo ": complete"; } |
unbuffered -vuc "$@" -- cat; }; show; show -b; show -bb; show -bbb

The unmodified stream currently is buffered as well with -b. This can be considered a bug. In future this might change.

The idea is:

With -b the typical "broken lines" which come from buffer flushes are understood. Note that there must be at least one complete line seen on each read.
With -bb a real buffering is done (up to 100000 byte or just a bit more), such that fragmented lines are joined together.
With -bbb there is no memory limit for buffering of lines. So this can eat up all of your memory, if there never is an LF seen.

(In the example above you do not see any difference between -bb and -bbb, as the line is too short.)

Known bugs

Hexdump is not yet UTF-8 aware. As it dumps all chars with bit 7 set, this might produce some buggy output on UTF-8 terminals, including swallowed LF. Hint: Pipe through less.

License

This Works is placed under the terms of the Copyright Less License, see file COPYRIGHT.CLL. USE AT OWN RISK, ABSOLUTELY NO WARRANTY.

This means, it is free as in free speech, free beer and free baby. (Babys usually have no Copyright on them, right?)

hilbix / unbuffered