sed -i -e '/^$/d' filename.txt

awk '/./{print > FILENAME}' yourfile

:g/^$/ d

#!/bin/bash

infile="infile.txt"
outfile="outfile.txt"
blocksize=1M
count=0

# process a "blocksize" block of text from the end of the file during each
# iteration of the loop, remove all the blank lines from it, and store in a temp
# file.  Remove an equivalent amount from the original file at the same time.

while [[ -s $infile ]]; do

		# read in a block of text, and truncate the original file by the same amount.
		# pads each end with an "x" also, which will be removed later.
		# this ensures that any newlines at the ends of the block are preserved.

		block=$( printf 'x' ; tail -c "$blocksize" "$infile" ; printf 'x' )
		truncate --size=-"$blocksize" "$infile"

		# pre-increment the counting variable for each block.
		(( count++ ))

		# "squeeze" all newlines in the block, and print it to a tempfile.
		# then remove the characters at the ends and print the result to
		# a temporary file, with $count in its name.
		# unwanted newlines from creeping back in.

		block=$( tr -s '\n' <<<"$block" )
		block=${block#x}
		block=${block%x}
		printf '%s' "$block" > "$outfile.$count"

done

# now cat all the tempfiles back together.
# since we started from the bottom we work
# from highest number to lowest.

while (( count )); do

	cat "$outfile.$count" >> "$outfile"
	rm "$outfile.$count"
	(( count-- ))

done

# Finish by adding a final newline back to the file (comment out if not desired).
echo >> "$outfile"

exit 0

# Some dd versions tell you the number of bytes
tr -s '\n' < file | dd of=file bs=1M conv=notrunc
0+140231 records in
0+140231 records out
1095007070 bytes .... # Number of bytes through the pipe (from dd)

# Otherwise, in Bash try this
tr -s '\n' < file | tee >(wc -c >&2) | dd of=file bs=1M conv=notrunc
1095007070  # Number of bytes through the pipe (from wc)
0+133329 records in
0+133329 records out

# Truncating the file to the new size:
dd if=/dev/null seek=1095007070 of=file bs=1
0+0 records in
0+0 records out