[webkit-reviews] review granted: [Bug 220668] Syscall param sendmsg(msg.msg_iov[1]) points to uninitialised byte(s) in IPC::Connection::sendOutgoingMessage : [Attachment 417777] Patch

[bugzilla-daemon at webkit.org]

Darin Adler <darin at apple.com> has granted Michael Catanzaro
<mcatanzaro at gnome.org>'s request for review:
Bug 220668: Syscall param sendmsg(msg.msg_iov[1]) points to uninitialised
byte(s) in IPC::Connection::sendOutgoingMessage
https://bugs.webkit.org/show_bug.cgi?id=220668

Attachment 417777: Patch

https://bugs.webkit.org/attachment.cgi?id=417777&action=review




--- Comment #22 from Darin Adler <darin at apple.com> ---
Comment on attachment 417777
  --> https://bugs.webkit.org/attachment.cgi?id=417777
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=417777&action=review

> Source/WebKit/Platform/IPC/unix/ConnectionUnix.cpp:80
> +    AttachmentInfo(const AttachmentInfo& info)
> +    {
> +	   memset(static_cast<void*>(this), 0, sizeof(*this));
> +	   m_type = info.m_type;
> +	   m_isNull = info.m_isNull;
> +	   m_size = info.m_size;
> +    }

This looks good to me.

> Source/WebKit/Platform/IPC/unix/ConnectionUnix.cpp:82
> +    AttachmentInfo(AttachmentInfo&& info)

On reflection, I think we can omit this function entirely. As long as we don’t
implement a move constructor it will use the copy constructor, which for this
class is fine. There’s no requirement when moving that the original object is
zeroed, just needs to be safe to destroy.

> Source/WebKit/Platform/IPC/unix/UnixMessage.h:59
> +    MessageInfo(MessageInfo&& info)

Same thought here. Don’t need this.

> Source/WebKit/Platform/IPC/unix/UnixMessage.h:69
> +    MessageInfo& operator=(const MessageInfo&) = default;
> +    MessageInfo& operator=(MessageInfo&&) = default;

Don’t need these.