Vim, 54, 52, 49 47 keystrokes

2wa0<esc>qqYp<c-a>6ldf @qq@q$dT]dd:%norm dwf{xwC;<CR>gg"0P

Explanation:

2wa0<esc>                     'Move 2 words forward, and insert a 0.
         qq                   'Start recording in register Q
           Yp                 'Duplicate the line
             <c-a>6l          'Increment the next number then move 6 spaces right
                    df        'Delete until the next space
                       @qq@q  'Recursively call this macro

Now our buffer looks like this:

int foo[0] = {4, 8, 15, 16, 23, 42};
int foo[1] = {8, 15, 16, 23, 42};
int foo[2] = {15, 16, 23, 42};
int foo[3] = {16, 23, 42};
int foo[4] = {23, 42};
int foo[5] = {42};
int foo[6] = {42};

Second half:

$                           'Move to the end of the line
 dT]                        'Delete back until we hit a ']'
    dd                      'Delete this whole line.
      :%norm         <CR>   'Apply the following keystrokes to every line:
             dw               
               f{x          '(f)ind the next '{' the delete it.
                  wC;       'Move a word, then (C)hange to the end of this line, 
                            'and enter a ';'

Now everything looks good, we just need to add the original array-declaration. So we do:

gg        'Move to line one
  "0P     'Print buffer '0' behind us. Buffer '0' always holds the last deleted line,
          'Which in this case is "int foo[6];"

Pyth, 44 bytes

++Khcz\]lJ:z"-?\d+"1"];"VJs[ecKd~hZ"] = "N\;

Test suite

Regular expression and string chopping. Not particularly clever.

Explanation:

++Khcz\]lJ:z"-?\d+"1"];"VJs[ecKd~hZ"] = "N\;
                                                Implicit: z = input()
    cz\]                                        Chop z on ']'
   h                                            Take string before the ']'
  K                                             Store it in K
 +                                              Add to that
         :z"-?\d+"1                             Find all numbers in the input
        J                                       Store them in J
       l                                        Take its length.
+                  "];"                         Add on "];" and print.
                       VJ                       For N in J:
                         s[                     Print the following, concatenated:
                            cKd                 Chop K on spaces.
                           e                    Take the last piece (array name)
                               ~hZ              The current interation number
                                  "] = "        That string
                                        N       The number from the input
                                         \;     And the trailing semicolon.

Retina, 108 104 100 69 bytes

Byte count assumes ISO 8859-1 encoding.

].+{((\S+ ?)+)
$#2];$1
+`((\w+\[).+;(\S+ )*)(-?\d+).+
$1¶$2$#3] = $4;

Beat this, PowerShell...

Code explanation

First line: ].+{((\S+ ?)+)

First, we need to keep the type, array name, and opening bracket (it saves a byte), so we don't match them. So we match the closing bracket, any number of characters, and an opening curly brace: ].+{. Then we match the number list. Shortest I've been able to find so far is this: ((\S+ ?)+). We match any number of non-space characters (this includes numbers, possible negative sign, and possible comma), followed by a space, that may or may not be there: \S+ ?. This group of characters is then repeated as many times as needed: (\S+ ?)+ and put into the large capturing group. Note that we don't match the closing curly brace or semicolon. Third line explanation tells why.

Second line: $#2];$1

Since we only matched a part of input, the unmatched parts will still be there. So we put the length of the list after the unmatched opening bracket: $#2. The replace modifier # helps us with that, as it gives us the number of matches a particular capturing group made. In this case capturing group 2. Then we put a closing bracket and a semicolon, and finally our whole list.

With input short array[] = {4, 3, 2, 1};, the internal representation after this replacing is:

short array[4];4, 3, 2, 1};

(note the closing curly brace and semicolon)

Third line: +`((\w+[).+;(\S+ )*)(-?\d+).+

This is a looped section. That means it runs until no stage in the loop makes a change to the input. First we match the array name, followed by an opening bracket: (\w+\[). Then arbitrary number of characters and a semicolon: .+;. Then we match the list again, but this time only numbers, which a space behind them (this excludes the last one): (\S+ )*. Then we take the last number in the list: (-?\d+) and any remaining characters behind it: .+.

Fourth line: $1¶$2$#3] = $4;

We then replace it with the array name and list followed by a newline: $1¶. Next, we put the array name, followed by the length of previously matched list, without the last element (essentially list.length - 1): $2$#3. Followed by a closing bracket and assigment operator with spaces, and this followed by the last element of our number list: ] = $4;

After first replacing, the internal representation looks like this:

short array[4];4, 3, 2, 
array[3] = 1;

Note that the closing curly brace and the semicolon disappeared, thanks to the .+ at the end of third line. After three more replacings, the internal representation looks like this:

short array[4];
array[0] = 4;
array[1] = 3;
array[2] = 2;
array[3] = 1;

Since there's nothing more to match by third line, fourth doesn't replace anything and the string is returned.

TL;DR: First we change up the int list format a bit. Then we take the last element of the list and the name, and put them after the array initialization. We do this until the int list is empty. Then we give the changed code back.

Try it online!

Python 3.6 (pre-release), 133

m,p=str.split,print;y,u=m(input(),'[');t,n=m(y);i=m(u[5:-2],', ')
l=len(i);p(t,n+f'[{l}];')
for x in range(l):p(n+f'[{x}] = {i[x]};')

Makes heavy use of f-strings.

Ungolfed version:

y, u = input().split('[')
t, n = y.split()
i = u[5:-2].split(', ')
l = len(i)
print(t, n + f'[{l}];')
for x in range(l):
    print(n + f'[{x}] = {i[x]};')

05AB1E, 52 50 47 bytes

Code:

… = ¡`¦¨¨ð-',¡©gr¨s«„];«,®v¹ð¡¦¬s\¨N"] = "y';J,

Uses CP-1252 encoding. Try it online!.

Pyth - 53 50 46 45 bytes

1 bytes saved thanks to @FryAmTheEggman.

+Jhcz\[+`]lKc<1ecz\{d\;j.es[ecJd`]kd\=dPb\;)K

Test Suite.

JavaScript (ES6), 100 bytes

(s,[t,n,...m]=s.match(/-?\w+/g))=>t+` ${n}[${m.length}];`+m.map((v,i)=>`
${n}[${i}] = ${v};`).join``

Since only the words are important, this works by just matching all the words in the original string, plus leading minus signs, then building the result. (I originally thought I was going to use replace but that turned out to be a red herring.)

C, 286 bytes

i,j,n;f(s,z,t)char*s,*z,*t;{z=s;for(j=i=0;z[i]-32;putchar(z[i++]));for(;z[i++]-91;t[j++]=z[i]);t[j]=0;for(n=1;z[i]-'}';)n+=z[i++]==44;printf(" %s%i]",t,n);for(i=j=0;z[i++]-'{';);for(;j<n;i++){printf(";\n%s%i] = ",t,j++);for(;z[i]-44&&z[i]-'}';i++)if(z[i]-32)putchar(z[i]);}printf(";");}

Ideone link

JavaScript ES6, 134 132 130 129 bytes

Saved 1 byte thanks to Neil.

x=>(m=x.match(/(\w+) (\w+).+{(.+)}/),m[1]+` `+(q=m[2])+`[${q.length-1}];
`+m[3].split`, `.map((t,i)=>q+`[${i}] = ${t};`).join`
`)

Python 2, 159 bytes

s=input().split()
t,n,v=s[0],s[1][:-2],''.join(s[3:])
a=v[1:-2].split(',')
print'%s %s[%d];'%(t,n,len(a))
for i in range(len(a)):print'%s[%d] = %s;'%(n,i,a[i])

Try it online

Thanks Kevin Lau for some golfing suggestions

Awk, 101 bytes

{FS="[^[:alnum:]_-]+";printf"%s %s[%d];\n",$1,$2,NF-3;for(i=3;i<NF;i++)printf$2"[%d] = %d;\n",i-3,$i}

More readably:

{
FS="[^[:alnum:]_-]+"
printf "%s %s[%d];\n", $1, $2, NF - 3
for (i=3; i < NF; i++)
    printf $2"[%d] = %d;\n", i-3, $i
}

• I set the field separator to everything except alphabets, digits, the underscore and -. So, the fields would be the type name, the variable name, and the numbers.
• The number of fields will be 1 (for the type) + 1 (for the name) + N (numbers) + 1 (an empty field after the trailing };). So, the size of the array is NF - 3.
• Then it's just printing a special line for the declaration, and looping over the numbers.
• I should assign FS either when invoking awk (using -F) or in a BEGIN block. In the interests of brevity, ….

PowerShell v2+, 114 105 bytes

$a,$b,$c,$d=-split$args-replace'\[]';"$a $b[$(($d=-join$d|iex|iex).length)];";$d|%{"$b[$(($i++))] = $_;"}

_{Currently beating the other non-esolangs, and closing in on beating Retina. Woo!}

Takes input string $args and -replaces the square bracket with nothing, then performs a -split on whitespace. We store the first bit into $a, the second bit into $b, the = into $c, and the array elements into $d. For the example below, this stores foo into $a and bar into $b, and all of the array into $d.

We then output our first line with "$a ..." and in the middle transform $d from a an array of strings of form {1,,2,,...100}; to a regular int array by -joining it together into one string, then running it through iex twice (similar to eval). We store that resultant array back in $d before calling the .length method to populate the appropriate number in between the [] in the output line.

We then send $d through a loop with |%{...}. Each iteration we output "$b..." with a counter variable $i encapsulated in brackets and the current value $_. The $i variable starts uninitialized (equivalent to $null) but the ++ will cast it to an int before the output, so it will start output at 0, all before incrementing $i for the next loop iteration.

All output lines are left on the pipeline, and output to the terminal is implicit at program termination.

Example

PS C:\Tools\Scripts\golfing> .\expand-a-c-array.ps1 "foo bar[] = {1, 2, 3, -99, 100};"
foo bar[5];
bar[0] = 1;
bar[1] = 2;
bar[2] = 3;
bar[3] = -99;
bar[4] = 100;

Ruby, 127 110 108 99 bytes

Anonymous function with a single argument as input. Full program, reads the input from STDIN. (If you pipe a file in, the trailing newline is optional.) Returns Prints the output string.

Took @TimmyD bragging about their solution beating all other non-esolangs as a challenge, and finally overcome the (at the time of writing) 114 byte Powershell solution they had posted. Cᴏɴᴏʀ O'Bʀɪᴇɴ's trick with splitting on ] and splicing the second half to get the numbers helped.

I need to use the splat operator more. It's so useful!

t,n,*l=gets.gsub(/[^ \w]/,"").split;i=-1
puts t+" #{n}[#{l.size}];",l.map{|e|n+"[#{i+=1}] = #{e};"}

Pip, 48 bytes

qR`(\S+)(. = ).(.+)}`{d^:k[b#d']';.n.b.,#d.c.d]}

Takes input from stdin and prints to stdout.

Full explanation pending, but essentially this does a regex replacement using capture groups and a callback function to construct the result.

bash, 133 129 bytes

read -a l
d="${l[@]:0:2}"
e=("${l[@]:3}")
echo "${d%?}${#e[@]}];"
for i in "${!e[@]}"
{
echo "${l[0]}[$i] = ${e[$i]//[!0-9]/};"
}

First attempt, sure its posible to get it shorter.

Julia, 154 148 144 137 134 bytes

f(s,c=match(r"(\w+ (\w+)).+{(.+)}",s).captures,x=split(c[3],", "),n=endof(x))=c[1]"[$n];\n"join([c[2]"[$i] = "x[i+1]";\n"for i=0:n-1])

This is a function that accepts a string and returns a string with a single trailing newline.

Ungolfed:

function f(s,
           c = match(r"((\w+ (\w+)).+{(.+)}", s).captures,
           x = split(c[3], ", "),
           n = endof(x))

    c[1] "[$n];\n" join([c[2] "[$i] = " x[i+1] ";\n" for i = 1:n-1])
end

We define c to be an array of capture groups where the first element is the type and array name, the second is the array's name, and the third is the list of values. We set x as the result of splitting the list of values into an array, then store n as the length of x. The output is constructed as the type, name and length string interpolated, combined with each definition line separated by newlines.

MATL, 68 64 58 bytes

'\w+'XX2:H#)XKxXIZc'['KnV'];'v!K"I2X)'['X@qV'] = '@g';'6$h

This isn't C, but it does use C-like sprintf function Nah, that was wasting 4 bytes.

Try it online!

          % Take input implicitly
'\w+'XX   % Find substrings that match regex '\w+'. Gives a cell array
2:H#)     % Split into a subarray with the first two substrings (type and name), and 
          % another with the rest (numbers that form the array)
XKx       % Copy the latter (numbers) into clipboard K. Delete it
XI        % Copy the former (type and name) into clipboard I
Zc        % Join the first two substrings with a space
'['       % Push this string
K         % Paste array of numbers
nV        % Get its length. Convert to string
'];'      % Push this string
v!        % Concatenate all strings up to now. Gives first line of the output
K"        % For each number in the array
  I2X)    %   Get name of array as a string
  '['     %   Push this string
  X@qV    %   Current iteration index, starting at 0, as a string
  '] = '  %   Push this string
  @g      %   Current number of the array, as a string
  ';'     %   Push this string
  5$h     %   Concatenate top 6 strings. This is a line of the output
          % Implicity end for each
          % Implicitly display

Python 3, 116 bytes

t,v,_,*l=input().split();v=v[:-1]+'%s]'
print(t,v%len(l)+';');i=0
for x in l:print(v%i,'= %s;'%x.strip('{,};'));i+=1

Splits the input into the type, name, and the list of numbers. After printing the array declaring, prints the elements by manually enumerating through the numbers, removing excess punctuation that attached to the first and last one.

A different approach in Python 2 came out to 122 bytes:

a,b=input()[:-2].split('] = {')
l=eval(b+',')
print a+`len(l)`+"];"
for y in enumerate(l):print a.split()[1]+'%s] = %s;'%y

The idea is to eval the list of numbers as a tuple, with a comma at the end so that a single number is recognized as a type. The enumerated list of numbers provides tuples to string-format in.

D, 197 bytes

import std.array,std.stdio;void main(){string t,n,e;readf("%s %s[] = {%s}",&t,&n,&e);auto v=e.replace(",","").split;writeln(t,' ',n,'[',v.length,"];");foreach(i,c;v)writeln(n,"[",i,"] = ",c,";");}

or ungolfed:

import std.array, std.stdio;

void main() {
    string type, name, elems;
    readf("%s %s[] = {%s}", &type, &name, &elems);

    // remove all commas before splitting the string into substrings
    auto vector = elems.replace(",","").split();

    // writeln is shorter than fln by 1 char when filled in
    writeln(type, ' ', name, '[', vector.length, "];");

    // print each element being assigned
    foreach(index, content; vector)
        writeln(name, "[", index, "] = ", content, ";");
}