Using GNU sed with the extended regex mode turned ON E

• Store all non semicolon records in hold.
• When we encounter the semicolon line, the merging process starts wherein the last portion of the pattern space is appended to the first portion of the pattern space and the first portion printed and stripped off. This continues till we exhaust the pattern space.

$ sed -Ee '/\n/ba
    /^[^;]+$/,/;/!b
    H;/;/!d;z;x;D;:a
    s/\n(.*\n)?[^;]+(;.*)/\2&/
    P;/\n.*\n/D;s/.*\n//
' file

$ perl -lne '$, = ";";
    push(@A,$_),next if !/;/;
    my $a = s/.*?;//r;
    print $_, $a for splice @A;
    print;
' file

Using GNU sed with the extended regex mode turned ON -E

• Store records in hold and wait for the semicolon record when the action begins.
• When we encounter the semicolon line, the merging process starts wherein the last portion of the pattern space(= semicolon line) is appended to the first portion of the pattern space and the first portion printed and stripped off. This continues till we exhaust the pattern space.

$ sed -Ee '/\n/ba
    H;/;/!d;z;x;D;:a
    s/\n(.*\n)?[^;]+(;.*)/\2&/
    P;/\n.*\n/D;s/.*\n//
' file

$ perl -lne '$, = ";";
    push(@A,$_),next if !/;/;
    my $a = s/.*?;//r;
    print $_, $a for splice @A;
    print;
' file