使用Posix分叉和管道的NativeCall代码不起作用

时间:2016-12-17 19:38:53

标签: c process posix perl6 nativecall

好的,所以我是Perl和Perl 6的新手。我以为我会看到我是否可以分叉和管道工作,但到目前为止一直无法做到。这是我的Perl 6代码:

use NativeCall;

# http://www.perlmonks.org/?node_id=989766
our sub c_close(int32) returns int32 is native is symbol('close') { * }
sub pipe(CArray[int32]) returns int32 is native { ... }
our sub c_read(int32, Str is encoded('utf8'), size_t) returns ssize_t is native is symbol('read') { *}
our sub c_write(int32, Str is encoded('utf8'), size_t) returns ssize_t is native is symbol('write') { *}
our sub c_wait(int32 is rw) is native is symbol('wait') { * }
sub c_puts(Str) is native is symbol("puts") { * }
sub waitpid(int32, Pointer, int32) returns int32 is native {*};

my @fd := CArray[int32].new;
#my ($child, $parent);
my $pok = pipe(@fd);
if ($pok == -1) { die "Pipe failed" ; }

sub fork() returns int32 is native { ... };

# See:
# https://rosettacode.org/wiki/Fork#Perl_6
my $pid = fork();
if ( $pid < 0) { die "Fork failed" ; }
if ( $pid == 0) {
        print "C: I am the child\n";
        if (c_close(@fd[1]) == -1) { die "Child couldn't close fd[1]" };
        my $msg_in = "";
        say "C: starting read";
        my $nread = c_read(@fd[0], $msg_in, 80);
        print "C: nread=$nread\n";
        print "C: message:$msg_in.\n";
        c_close(@fd[0]);
} else {
        print "P: I am the parent of $pid\n";
        if (c_close(@fd[0]) == -1) { die "Parent couldn't close fd[0]"; } ;
        my $msg  = "Hello from parent";
        my $len =  $msg.encode('utf8').bytes + 1;
        print "P: test put string: ";
        c_puts($msg);
        #print "P: len=$len\n";
        my $nwritten =c_write(@fd[1], $msg, $len);
        print "P: len $len, wrote $nwritten\n";
        say "P: Finished writing";
        c_close(@fd[1]);
        #my $null= 0;
        #c_wait($null);
        my $stat_loc;
        waitpid($pid, $stat_loc ,0);
}

这是运行它的结果:

P: I am the parent of 25809
C: I am the child
C: starting read
C: nread=-1
C: message:.
P: test put string: Hello from parent
P: len 18, wrote -1
P: Finished writing

似乎c_read()函数由于某种原因没有阻塞,AFAIK是不可能的。并不是说我对分叉很了解。

任何想法是什么修复?

答案

2016年12月19日更新

感谢@timotimo,我得到了一个有效的解决方案。似乎我的努力可以改进。例如,我认为它不适用于UTF-8。 Anyhoo,至少“它有效”。

use NativeCall;

# http://www.perlmonks.org/?node_id=989766
our sub c_close(int32) returns int32 is native is symbol('close') { * }
our sub c_fork() returns int32 is native is symbol('fork') { ... };
our sub c_pipe(CArray[int32]) returns int32 is native is symbol('pipe') { ... }
our sub c_puts(Str) is native is symbol("puts") { * }
our sub c_read(int32, CArray[uint8], size_t) returns ssize_t is native is symbol('read') { *}
our sub c_wait(int32 is rw) is native is symbol('wait') { * }
our sub c_waitpid(int32, Pointer, int32) returns int32 is native is symbol('waitpid') {*};
our sub c_write(int32, Str is encoded('utf8'), size_t) returns ssize_t is native is symbol('write') { *}


my @fd := CArray[int32].new(0, 0);
my $pok = c_pipe(@fd);
if ($pok == -1) { die "Pipe failed" ; }


# See:
# https://rosettacode.org/wiki/Fork#Perl_6
my $pid = c_fork();
if ( $pid < 0) { die "Fork failed" ; }
if ( $pid == 0) {
        print "C: I am the child\n";
        if (c_close(@fd[1]) == -1) { die "Child couldn't close fd[1]" };
        my uint8 $b0 = 0;
        my @buf := CArray[uint8].new( $b0 xx 80);
        say "C: starting read";
        my $nread = c_read(@fd[0], @buf, 80);
        print "C: nread=$nread\n";
        my $msg = "";
        for (0..$nread-1) -> $i { $msg = $msg ~ chr(@buf[$i]); } ;
        print "C: message:$msg.\n";
        c_close(@fd[0]);
} else {
        print "P: I am the parent of $pid\n";
        if (c_close(@fd[0]) == -1) { die "Parent couldn't close fd[0]"; } ;
        my $msg  = "Hello from parent";
        my $len =  $msg.encode('utf8').bytes;
        print "P: test put string: ";
        c_puts($msg);
        my $nwritten =c_write(@fd[1], $msg, $len);
        print "P: len $len, wrote $nwritten\n";
        say "P: Finished writing";
        c_close(@fd[1]);
        my $stat_loc;
        c_waitpid($pid, $stat_loc ,0);
}

现在按预期输出:

P: I am the parent of 22448
C: I am the child
P: test put string: Hello from parent
C: starting read
P: len 17, wrote 17
P: Finished writing
C: nread=17
C: message:Hello from parent.

我创建了gist,并根据需要修改了解决方案。

1 个答案:

答案 0 :(得分:8)

你的问题完全不同。

您创建了CArray,但实际上并没有为管道要写入的两个整数腾出空间。写入进入who-know-where,而你的@fd只有内容[0, 0],所以你在读写时得到了BADF(Bad File Descriptor),所以他们立即返回。 / p> 每当你使用posix api时,

strace -f都是一个很棒的工具。这是什么给了我正确的想法。

以下是使@fd工作的所需代码:

my @fd := CArray[int32].new(0, 0);
pipe made
here's the pipe fds
17
18
P: I am the parent of 13943
C: I am the child
C: starting read
P: test put string: Hello from parent
P: len 18, wrote 18
P: Finished writing
C: nread=18
C: message:.

PS:消息没有正确写入,因为c_read的Str参数不像你期望的那样工作。您将不得不再次执行相同的CArray,给它正确的大小(通过分配0 xx $size或执行@result[$size + 1] = 0),然后您将必须将其解码为utf8或latin1或者什么 - 具备-你。