scanf behaviour is strange

[yuantailing]

scanf behaviour is strange.

main.cpp: read at most 200 numbers and print them.

#include <cstdio>

using namespace std;

int main() {
    for (int i = 0; i < 200; i++) {
        int x;
        int ret = scanf("%d", &x);
        if (ret == EOF)
            break;
        printf("%d ", x);
        if ((i + 1) % 10 == 0)
            printf("\n");
    }
    return 0;
}

Input (in.txt):

0 1 2 3 4 5 6 7 8 9
10 11 12 13 14 15 16 17 18 19
20 21 22 23 24 25 26 27 28 29
30 31 32 33 34 35 36 37 38 39
40 41 42 43 44 45 46 47 48 49
50 51 52 53 54 55 56 57 58 59
60 61 62 63 64 65 66 67 68 69
70 71 72 73 74 75 76 77 78 79
80 81 82 83 84 85 86 87 88 89
90 91 92 93 94 95 96 97 98 99
100 101 102 103 104 105 106 107 108 109
110 111 112 113 114 115 116 117 118 119
120 121 122 123 124 125 126 127 128 129
130 131 132 133 134 135 136 137 138 139
140 141 142 143 144 145 146 147 148 149

Compile with default options:

user@host:~$ em++ main.cpp

Output is incorrect when using bash redirection:

user@host:~$ node a.out.js <in.txt
0 1 2 3 4 5 6 7 8 9
10 11 12 13 14 15 16 17 18 19
20 21 22 23 24 25 26 27 28 29
30 31 32 33 34 35 36 37 38 39
40 41 42 43 44 45 46 47 48 49
50 51 52 53 54 55 56 57 58 59
60 61 62 63 64 65 66 67 68 69
70 71 72 73 74 75 76 77 78 79
80 81 82 83 84 85 86 87 880 1
2 3 4 5 6 7 8 9 10 11
12 13 14 15 16 17 18 19 20 21
22 23 24 25 26 27 28 29 30 31
32 33 34 35 36 37 38 39 40 41
42 43 44 45 46 47 48 49 50 51
52 53 54 55 56 57 58 59 60 61
62 63 64 65 66 67 68 69 70 71
72 73 74 75 76 77 78 79 80 81
82 83 84 85 86 87 880 1 2 3
4 5 6 7 8 9 10 11 12 13
14 15 16 17 18 19 20 21 22 23

Output is correct when using pipe:

user@host:~$ cat in.txt | node a.out.js
0 1 2 3 4 5 6 7 8 9
10 11 12 13 14 15 16 17 18 19
20 21 22 23 24 25 26 27 28 29
30 31 32 33 34 35 36 37 38 39
40 41 42 43 44 45 46 47 48 49
50 51 52 53 54 55 56 57 58 59
60 61 62 63 64 65 66 67 68 69
70 71 72 73 74 75 76 77 78 79
80 81 82 83 84 85 86 87 88 89
90 91 92 93 94 95 96 97 98 99
100 101 102 103 104 105 106 107 108 109
110 111 112 113 114 115 116 117 118 119
120 121 122 123 124 125 126 127 128 129
130 131 132 133 134 135 136 137 138 139
140 141 142 143 144 145 146 147 148 149

My environment:

user@host:~$ em++ --version
emcc (Emscripten gcc/clang-like replacement) 1.37.39 (commit 63ccba9f6307da3fcccaa9324252da54f4cba505)
Copyright (C) 2014 the Emscripten authors (see AUTHORS.txt)
This is free and open source software under the MIT license.
There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
user@host:~$ node --version
v8.9.1
user@host:~$ bash --version
GNU bash, version 4.3.11(1)-release (x86_64-pc-linux-gnu)
Copyright (C) 2013 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>

This is free software; you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.

[kripken]

I am surprised this works at all - we don't seem to have any special code to handle stdin in src/shell.js - it's something that needs to be done manually (to define Module['stdin']). So I must be missing something here.

(I'm also unsure about the bash differences between pipes and redirection, I would have guessed both work as pipes internally...)

[stale]

This issue has been automatically marked as stale because there has been no activity in the past year. It will be closed automatically if no further activity occurs in the next 7 days. Feel free to re-open at any time if this issue is still relevant.

[yuantailing]

The problem still exists in current version.

emcc (Emscripten gcc/clang-like replacement) 1.38.45 (commit 252410a)

[bvibber]

Underlying code seems to use get_char which on Node on Linux/Mac is opening /dev/stdin for read, reading up to 256 bytes, and buffering those for read. There then seems to be some different behavior between node a.out.js < in.txt and cat in.txt | node a.out.js at the lower levels of the stack -- the redirection case feeds in a full 256 bytes on the first read, then the second read reads the first 256 bytes again for a total of four 256-byte reads. We only scan through the first and second, but both return the first 256 bytes given! On the pipe case, it correctly fetches 256 bytes then the following 234 bytes, and scans the expected input data.

[bvibber]

I can repro with this extracted bit on node:

var fs = require('fs');

function get_chars() {
    // we will read data by chunks of BUFSIZE
    var BUFSIZE = 256;
    var buf = Buffer.alloc ? Buffer.alloc(BUFSIZE) : new Buffer(BUFSIZE);
    var bytesRead = 0;

    var isPosixPlatform = (process.platform != 'win32'); // Node doesn't offer a direct check, so test by exclusion

    var fd = process.stdin.fd;
    if (isPosixPlatform) {
        // Linux and Mac cannot use process.stdin.fd (which isn't set up as sync)
        var usingDevice = false;
        try {
            fd = fs.openSync('/dev/stdin', 'r');
            usingDevice = true;
        } catch (e) {}
    }

    try {
        bytesRead = fs.readSync(fd, buf, 0, BUFSIZE, null);
    } catch(e) {
    // Cross-platform differences: on Windows, reading EOF throws an exception, but on other OSes,
    // reading EOF returns 0. Uniformize behavior by treating the EOF exception to return 0.
    if (e.toString().indexOf('EOF') != -1) bytesRead = 0;
        else throw e;
    }

    if (usingDevice) { fs.closeSync(fd); }
    if (bytesRead > 0) {
        result = buf.slice(0, bytesRead).toString('utf-8');
    } else {
        result = null;
    }
    return result;
}

for (var i = 0; i < 2; i++) {
    let s = get_chars();
    console.log({s});
}

Second get_chars() call buffers up the beginning of input again instead of the second chunk as expected, but only when using file redirection and not when using a pipe.

[bvibber]

I get the same 'restarting' behavior with a C program on Linux, so I think the code is just not using /dev/stdin in the way it's intended to be used?

[sbc100]

If i remove the isPosixPlatform this seems to work as expected.. so perhaps we can simply remove that workaround?