From: Eric Christopherson <echristopherson@...>
Date: 2010-02-22T09:15:44+09:00
Subject: Re: Difference between system() versus exec Kernel methods

On Sun, Feb 21, 2010 at 5:58 PM, Walton Hoops <walton@vyper.hopto.org> wrote:
> On 2/21/2010 4:38 PM, Alex DeCaria wrote:
>>
>> That's well and good.  I've already read the documentation.  But, based
>> on the documentation I would expect to see only a single process running
>> after calling the exec method.  So, why are there two?
>>
>> I'm obviously not reading the documentation correctly, and that's why I
>> need someone who can explain it.
>>
>> --Alex
>>
>
> This line from the docs explains it:
>
>     In MSDOS environments, the command is executed in
>     a subshell; otherwise, one of the +exec(2)+ system calls is used,
>     so the running command may inherit some of the environment of the
>     original program (including open file descriptors).
>
> The trick is, exec uses the underlying OSes exec.  In Linux, that does as
> you
> expect and simply replaces the running process in memory.  In Windows
> however,
> there is no equivalent to the Linux exec, so we pretend. Ruby execs the
> specified command (via the Windows version of exec which starts a second
> process)
> and exits once it's done, thus there are two processes.

Sounds like the doc should be updated to mention Windows.

>
> To demonstrate this try running the following:
>
> exec "echo boo"
> puts "I hope I don't get here"
>
>
> Notice that the puts statement never gets run.