<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=UTF-8" http-equiv="Content-Type">
<title></title>
</head>
<body bgcolor="#ffffff" text="#000000">
Olly Betts wrote:
<blockquote cite="mid20060421035405.GI20016@survex.com" type="cite">
<pre wrap="">On Thu, Apr 20, 2006 at 05:11:14PM +0200, Daniel M?nard wrote:
</pre>
<blockquote type="cite">
<pre wrap="">BTW, you remember me a problem I had with my wrapper and that I have not
investigated : I had to explicitely release the object (php should do it
automatically). For example, with a WritableDatabase, the lock file was not
removed if I didn't write "$db=null".
I will have to test the new wrapper to check that we don't have a similar
problem.
</pre>
</blockquote>
<pre wrap=""><!---->
I just tested this with the latest version (8) and I don't get the lock
file left behind with a quartz database. However, I don't if I use the
flat file API either...
</pre>
</blockquote>
Can we do anything to be more sure that all the destructors are being
called?<br>
<br>
Olly, I'm still concerned about the case where wrapped objects are
referenced by other wrapped objects (meaning at the c++ xapian level)<br>
<br>
I tried to make swig-php be able to call destructors, or not, depending
on whether or not we constructed or merely wrapped objects, I hope I
got this right, but I realise that sometimes even though we created the
object it may get referenced from another object, and I recall that
xapian uses smart reference counting pointers.<br>
<br>
Thus, if the wilful destructor stage is working in shutdown, some
objects might get twice destroyed, (and he we see how I've forgotten so
much) unless php-swig's wilful destructor stage only releases a ref
count after all.<br>
<br>
Sam<br>
<br>
</body>
</html>