Nekomata, 7 bytes

xSᵖ{@∑=

Attempt This Online!

xSᵖ{@∑=
 S          Find a subset of
x           [0, 1, ..., len(first input) - 1]
  ᵖ{        such that
     ∑      the sum of
    @       the corresponding elements of the first input
      =     equals the second input

By default, Nekomata prints all possible solutions.

APL(NARS), 49 chars, 98 bytes

{∨/b←⍺=+/¨{∊⍵⊂w}¨n←{⍵⊤⍨k⍴2}¨⍳¯1+2*k←≢w←⍵:⍸¨b/n⋄⍬}

1-indexed; test:

  f←{∨/b←⍺=+/¨{∊⍵⊂w}¨n←{⍵⊤⍨k⍴2}¨⍳¯1+2*k←≢w←⍵:⍸¨b/n⋄⍬}
  ⎕fmt 8 f 1 2 1 5
┌2──────────────┐
│┌3────┐ ┌3────┐│
││2 3 4│ │1 2 4││
│└~────┘ └~────┘2
└∊──────────────┘
  ⎕fmt   14.8  f  4.8 9.5 2.7 11.12 10
┌1────┐
│┌2──┐│
││1 5││
│└~──┘2
└∊────┘
  ⎕fmt 17 f 7, 8, 9, ¯10, 20, 27
┌3──────────────────┐
│┌2──┐ ┌2──┐ ┌3────┐│
││4 6│ │2 3│ │1 4 5││
│└~──┘ └~──┘ └~────┘2
└∊──────────────────┘
  ⎕fmt 7 f 1 2 3
┌0┐
│0│
└~┘

comment:

{∨/b←⍺=+/¨{∊⍵⊂w}¨n←{⍵⊤⍨k⍴2}¨⍳¯1+2*k←≢w←⍵:⍸¨b/n⋄⍬}
                             ⍳¯1+2*k←≢w←⍵         copy ⍵ in w, len(⍵) in k, return 1..2^(k-1) 
                 n←{⍵⊤⍨k⍴2}¨                     traslate in binary each element of  1..2^(k-1) and assign the result to n
          {∊⍵⊂w}¨                                for each binary element of n return the elemets of ⍵ in the place where there are the 1s
    b←⍺=+/¨                                       sum them and see if the sum is ⍺, that binary array saved in b
  ∨/                                     :⍸¨b/n   if or/b, get all the elements of n that are 1s for array b, and calculate each all indexs
                                               ⋄⍬ else return Zilde as void set

Brachylog, 18 15 bytes

hiᶠ⊇Shᵐ+~t?∧Stᵐ

Try it online!

-3 bytes because it now works as a generator. (It's probably possible to golf more, but working around the need to use indices is awkward.)

    S              The variable S
   ⊇               is a sublist of
  ᶠ                the list of all
 i                 pairs [element, index] from
h                  the first element of
                   the input;
     hᵐ            the first elements of each pair
       +           add up to
        ~t         the last element of
          ?        the input
           ∧       which isn't necessarily
            S      S,
             tᵐ    from which the last elements of each pair
                   are output.

Pyth, 11 bytes

fqvzs@LQTyU

Try it online here, or verify all the test cases at once here.

fqvzs@LQTyUQ   Implicit: Q=input 1 (list of numbers), z=input 2 (target value, as string)
               Trailing Q inferred
          UQ   Generate range [0-<length of Q>)
         y     Powerset of the above
f              Keep elements of the above, as T, when the following is truthy:
      L T        Map elements of T...
     @ Q         ... to the indicies in Q
    s            Take the sum
 q               Is the above equal to...
  vz             z as an integer
               Implicit print of the remaining results

J, 32 31 bytes

(=1#.t#])<@I.@#t=.1-[:i.&.#.1"0

Try it online!

                  1-[:i.&.#.1"0         Make a list of all masks
                                        for the input list. We flip the bits
                                        to turn the unnecessary (0...0)         
                                        into (1...1) that would be missing.
                                        Define it as t.

(=1#.t#])                               Apply the masks, sum and
                                        compare with the target

         <@I.@#                         Turn the matching masks into 
                                        lists of indices

Husk, 10 bytes

ηλfo=¹ṁ⁰tṖ

1-indexed. Try it online!

Explanation

ηλfo=¹ṁ⁰tṖ  Inputs are a number n (explicit, accessed with ¹) and a list x (implicit).
η           Act on the incides of x
 λ          using this function:
         Ṗ   Take all subsets,
        t    remove the first one (the empty subset),
  f          and keep those that satisfy this:
      ṁ⁰      The sum of the corresponding elements of x
   o=¹        equals n.

This uses the latest addition to Husk, η (act on indices). The idea is that η takes a higher order function α (here the inline lambda function) and a list x, and calls α on the indexing function of x (which is ⁰ in the above program) and the indices of x. For example, ṁ⁰ takes a subset of indices, maps indexing to x over them and sums the results.

JavaScript (ES6), 96 bytes

Takes input in currying syntax (list)(target).

a=>s=>a.reduce((b,_,x)=>[...b,...b.map(y=>[...y,x])],[[]]).filter(b=>!b.reduce((p,i)=>p-a[i],s))

Test cases

This would fail on the 2nd test case if 4.8 and 10 were swapped because of an IEEE 754 precision error -- i.e. 14.8 - 4.8 - 10 == 0 but 14.8 - 10 - 4.8 != 0. I think this is fine, although there may be a more relevant reference somewhere in meta.

let f =

a=>s=>a.reduce((b,_,x)=>[...b,...b.map(y=>[...y,x])],[[]]).filter(b=>!b.reduce((p,i)=>p-a[i],s))

console.log(JSON.stringify(f([1, 2, 1, 5])(8)))
console.log(JSON.stringify(f([4.8, 9.5, 2.7, 11.12, 10])(14.8)))
console.log(JSON.stringify(f([7, 8, 9, -10, 20, 27])(17)))
console.log(JSON.stringify(f([1, 2, 3])(7)))

Commented

a => s =>                 // given an array a[] of length N and an integer s
  a.reduce((b, _, x) =>   // step #1: build the powerset of [0, 1, ..., N-1]
    [ ...b,               //   by repeatedly adding to the previous list b[]
      ...b                //   new arrays made of:
      .map(y =>           //     all previous entries stored in y[]
        [...y, x]         //     followed by the new index x
      )                   //   leading to:
    ],                    //   [[]] -> [[],[0]] -> [[],[0],[1],[0,1]] -> ...
    [[]]                  //   we start with a list containing an empty array
  )                       // end of reduce()
  .filter(b =>            // step #2: filter the powerset
    !b.reduce((p, i) =>   //   keeping only entries b
      p - a[i],           //     for which sum(a[i] for i in b)
      s                   //     is equal to s
    )                     //   end of reduce()
  )                       // end of filter()

Wolfram Language (Mathematica), 43 bytes

1-indexed.

Pick[s=Subsets;s@Range@Tr[1^#],Tr/@s@#,#2]&

Try it online!

Perl 6, 45 bytes

->\a,\b{grep {a[$_].sum==b},^a .combinations}

Test it

Expanded:

->
  \a, # input list
  \b, # input target
{

  grep

  {
      a[ $_ ].sum # use the list under test as indexes into ｢a｣
    ==
      b
  },

  ^a              # Range upto ｢a｣ (uses ｢a｣ as a number)
  .combinations   # get all of the combinations
}

R, 85 84 bytes

function(l,k){N=combn
o={}
for(i in I<-1:sum(l|1))o=c(o,N(I,i,,F)[N(l,i,sum)==k])
o}

Try it online!

1-indexed.

combn usually returns a matrix, but setting simplify=F returns a list instead, allowing us to concatenate all the results together. combn(I,i,,F) returns all combinations of indices, and we take N(l,i,sum)==k as an index into that list to determine those which equal k.

Jelly, 11 bytes

ị³S=
JŒPçÐf

Try it online!

1-indexed. 4 bytes spent on returning indices rather than just the elements themselves.

-1 byte thanks to user202729
-1 byte thanks to Jonathan Allan

Clean, 104 102 98 bytes

import StdEnv
@n=[[]:[[i:b]\\i<-[0..n-1],b<- @i]]
?l n=[e\\e<-tl(@(length l))|sum(map((!!)l)e)==n]

Try it online!

Haskell, 76 bytes

x#t=[i|i<-tail$concat<$>mapM(\z->[[],[z]])[0..length x-1],t==sum[x!!k|k<-i]]

Try it online!

Python 2, 110 bytes

lambda a,n:[b for b in[[i for i in range(len(a))if j&1<<i]for j in range(2**len(a))]if sum(a[c]for c in b)==n]

Try it online!

Japt, 14 bytes

m, à f_x!gU ¥V

Test it online!

How it works

m, à f_x!gU ¥V   Implicit: U = input array, V = target sum
m,               Turn U into a range [0, 1, ..., U.length - 1].
   à             Generate all combinations of this range.
     f_          Filter to only the combinations where
       x           the sum of
        !gU        the items at these indices in U
            ¥V     equals the target sum.
                 Implicit: output result of last expression

Python 3, 144 bytes

lambda a,t:[[e for e,_ in x]for r in range(len(a))for x in combinations(list(enumerate(a)),r+1)if sum(y for _,y in x)==t]
from itertools import*

Try it online!

0-indexed. 44 bytes spent on returning indices rather than just the elements themselves.