From: Kevin Lewis
Date: June 30 2009 4:01pm
Subject: Re: IndexRootPage::splitIndexPage
List-Archive: http://lists.mysql.com/falcon/775
Message-Id: <4A4A36EF.4010700@sun.com>
MIME-Version: 1.0
Content-Type: text/plain; format=flowed; charset=ISO-8859-1
Content-Transfer-Encoding: 8BIT

LarsErik,

Yes, your analysis seems to be correct in that the return true is a catch22 that cannot happen.  It will only return true if the recursive call returns true.

Lars-Erik Bjørk wrote:
> Hi all!
> 
> I have been looking at indexRootPage::splitIndexPage, and a small part 
> of it seems a little strange to me. It will not provoke a bug, but it is 
> confusing when reading it.
> 
> The method is declared to return a bool. When splitting an index page, 
> the method will recursively call itself if the parent page also needs to 
> be split. The recursion will stop when the parent page does not need to 
> be split, or the root page is split. In this case, it will return false.
> 
> The recursion looks like this
> 
> 
> ...
> 
> if (splitIndexPage (dbb, indexId, parentBdb, transId, result, &splitKey,
>                               splitPageNumber, (parentBdb->pageNumber == 
> rootPageNumber)))
>     return true;
> 
> ...
> 
> However, this is the only place the method returns true, so therefore it 
> can actually never return true.
> 
> The return value, which will always be false is also used in a branch in 
> IndexRootPage::addIndexEntry.
> 
> Do you agree with this? It is ok if I rewrite this to be a void method 
> instead? It will improve the readability, and we will avoid the 
> possibility that someone will use the return value in the future.
> 
> Best,
> 
> Lars-Erik
> 
>