fixup_after_fork (NULL)

fixup_after_fork (NULL)

[Ken Brown]

fhandler_socket_wsock::fixup_after_exec and 
fhandler_socket_unix::fixup_after_exec call fixup_after_fork (NULL), which would 
seem to lead to calls to DuplicateHandle with first argument NULL.  I can't find 
any documentation that says this is legal.  What do these calls do?

Ken

Re: fixup_after_fork (NULL)

[Corinna Vinschen]

[-- Attachment #1: Type: text/plain, Size: 1160 bytes --]

On Feb 27 17:24, Ken Brown wrote:
> fhandler_socket_wsock::fixup_after_exec and
> fhandler_socket_unix::fixup_after_exec call fixup_after_fork (NULL), which
> would seem to lead to calls to DuplicateHandle with first argument NULL.  I
> can't find any documentation that says this is legal.  What do these calls
> do?

They fail with ERROR_INVALID_HANDLE.

I checked the commit history and it turned out that this has been
introduced with commit 528060195ca6c back in 2002.

In fhandler_socket_wsock it has no negative impact.  fixup_after_exec
calls fixup_after_fork *only* on need_fixup_before(), that is, if the
handle was some of the special LSP handles, so the socket handle has to
be created by WSADuplicateSocket/WSASocket.  In that case the parent
handle isn't needed.

Same in case of fhandler_pty_slave.  Its fixup_after_fork doesn't
require the parent handle at all.

In case of fhandler_socket_unix, this was most likely a straight
copy/paste or developer-didn't-actually-think-about-it-error.
I fear it was the latter.  I have to start thinking about it...


Thanks,
Corinna

-- 
Corinna Vinschen
Cygwin Maintainer

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: fixup_after_fork (NULL)

[Corinna Vinschen]

[-- Attachment #1: Type: text/plain, Size: 1292 bytes --]

On Feb 28 13:41, Corinna Vinschen wrote:
> On Feb 27 17:24, Ken Brown wrote:
> > fhandler_socket_wsock::fixup_after_exec and
> > fhandler_socket_unix::fixup_after_exec call fixup_after_fork (NULL), which
> > would seem to lead to calls to DuplicateHandle with first argument NULL.  I
> > can't find any documentation that says this is legal.  What do these calls
> > do?
> 
> They fail with ERROR_INVALID_HANDLE.
> 
> I checked the commit history and it turned out that this has been
> introduced with commit 528060195ca6c back in 2002.
> 
> In fhandler_socket_wsock it has no negative impact.  fixup_after_exec
> calls fixup_after_fork *only* on need_fixup_before(), that is, if the
> handle was some of the special LSP handles, so the socket handle has to
> be created by WSADuplicateSocket/WSASocket.  In that case the parent
> handle isn't needed.
> 
> Same in case of fhandler_pty_slave.  Its fixup_after_fork doesn't
> require the parent handle at all.
> 
> In case of fhandler_socket_unix, this was most likely a straight
> copy/paste or developer-didn't-actually-think-about-it-error.
> I fear it was the latter.  I have to start thinking about it...

This should be fixed now.  Thanks for noticing!


Corinna

-- 
Corinna Vinschen
Cygwin Maintainer

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]