fix file descriptor leak in example http server code

[adamyi]

Originally the code only closed the fd for the connection but not for the file.

I'm running this code in prod and it ran out of fds fairly quickly. This fix makes it more stable :)

[adamyi]

Hey @ajyoon, just a friendly ping on this :)

[ajyoon]

apologies for my delay in patching in this very important production server fix. It seems correct enough to me, but I have a commentary question

[ajyoon]

The comment above says it moves to cell 51, but this comment here says it's 49?

[kailando]

That is confusing.

[adamyi]

@ajyoon @kailando
I am very sorry about the delay in fixing the comment for this very important production server fix, and the confusion this has caused.

I verified that the comment was wrong, but the code was correct.

According to sendfile call:

  Currently in cell 48
  ++++++++++++++++++++++++++++++++++++++++ > sendfile() == code 40
  ++++ > Arg count: 4

  Arg 1: int out_fd
  (0) > Arg type: Normal
  +   > Arg len:  1
  Arg contents are currently stored in cell 4; go fetch it
  <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
  [ -
    >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
    +
    <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
  ]
  >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
  >

  Arg 2: int in_fd
  (0) > Arg type: Normal
  +   > Arg len:  1
  Arg contents are currently stored in cell 65; go fetch it
  >>>>>>>>>>
  [ -
    <<<<<<<<<<
    +
    >>>>>>>>>>
  ]
  <<<<<<<<<<
  >

Memory layout is that:

Cell 48: sendfile code
Cell 49: arg count 4
Cell 50: arg type
cell 51: arg len 1
Cell 52: out_fd
Cell 53: arg_type
cell 54: arg len 1
Cell 55: in_fd

Our previous close call ended at cell 48. Indeed, moving it right by 3 would move to 51, and we're emptying out 51-54 so that 55 can be the first argument for close syscall. After close syscall, we shift left by 3, and would indeed return back to 48.