Fixed WASI isatty

[syrusakbary]

Description

Current WASI implementation returns a wrong response when libc isatty is used for stdin, stdout or stderr.

This PR fixes it.

Review

• Add a short description of the the change to the CHANGELOG.md file

[syrusakbary]

This is a C file I created to debug the issue:

#include <wasi/core.h>
#include <__errno.h>

int __isatty_check(int fd) {
    __wasi_fdstat_t statbuf;
    int r = __wasi_fd_fdstat_get(fd, &statbuf);
    if (r != 0) {
        errno = r;
        return 0;
    }

    printf("fs_filetype: %i %i (=%i), rightbase: %i %i\n", statbuf.fs_filetype, __WASI_FILETYPE_CHARACTER_DEVICE, statbuf.fs_filetype != __WASI_FILETYPE_CHARACTER_DEVICE, statbuf.fs_rights_base, statbuf.fs_rights_base & (__WASI_RIGHT_FD_SEEK | __WASI_RIGHT_FD_TELL)); fflush(stdout);
    // A tty is a character device that we can't seek or tell on.
    if (statbuf.fs_filetype != __WASI_FILETYPE_CHARACTER_DEVICE ||
        (statbuf.fs_rights_base & (__WASI_RIGHT_FD_SEEK | __WASI_RIGHT_FD_TELL)) != 0) {
        errno = __WASI_ENOTTY;
        return 0;
    }
    return 1;

}

[syrusakbary]

bors try

[bors]

try

Build succeeded

• continuous-integration/travis-ci/push
• wasmerio.wasmer

[MarkMcCaskey]

Nice fix!

We'll probably want to update the WasiFile trait so that our file swapping API works properly with this too, but that can be done as a follow up.

Edit: to clarify I mean pushing the "is this a TTY?" logic into the trait so that users can control the output here and that stdout, stdin, and stderr continue to return the correct results even when they're not FDs 0, 1, 2

[syrusakbary]

bors r+

[bors]

Build succeeded

• continuous-integration/travis-ci/push
• wasmerio.wasmer