Look, I don&#39;t understand why this is so hard for the project to understand:<div><br></div><div>Certain Windows functions are BROKEN. Certain Windows algorithms make NO SENSE. Other things are BADLY OPTIMIZED. Other things make WRONG ASSUMPTIONS.</div>

<div><br></div><div>But these set of behaviors and design failures are what billions of customers and pieces of software DEPEND ON. Microsoft cannot change the STACK LAYOUT of IE due to fears that BHOs which look up the file handle up the stack will break. They have SHIMS that LIE about the stack order to make those BHOs work, even as the stack changes.</div>

<div><br></div><div>Microsoft, on this function, writes,</div><div><br></div><div>&quot;<meta charset="utf-8"><span class="Apple-style-span" style="font-family: &#39;Segoe UI&#39;, Verdana, Arial; font-size: 13px; "><a href="" class="LW_CollapsibleArea_TitleAhref" title="Collapse" id="671f5c2c-47f3-4bdc-b9d5-dac2cb92b5df" style="outline-style: none; outline-width: initial; outline-color: initial; color: rgb(19, 100, 196); text-decoration: none; display: inline !important; "><span class="LW_CollapsibleArea_Title" style="font-family: &#39;Segoe UI&#39;, Verdana, Arial; font-size: 1.538em; color: rgb(63, 82, 156); font-weight: bold; word-break: break-all; word-wrap: break-word; padding-left: 6px; ">Remarks</span></a></span><div class="LW_CollapsibleArea_TitleDiv" style="font-family: &#39;Segoe UI&#39;, Verdana, Arial; margin-top: 19px; margin-bottom: 19px; ">

<div><div class="LW_CollapsibleArea_HrDiv" style="font-size: 13px; padding-top: 12px; "><hr class="LW_CollapsibleArea_Hr" style="border-bottom-color: rgb(229, 229, 229); border-bottom-width: 1px; border-bottom-style: solid; border-top-style: none; border-top-width: initial; border-top-color: initial; border-right-style: none; border-right-width: initial; border-right-color: initial; color: rgb(229, 229, 229); margin-left: 4px; ">

</div></div></div><div class="sectionblock" id="671f5c2c-47f3-4bdc-b9d5-dac2cb92b5df_c" style="padding-left: 15px; display: block; clear: both; "><font class="Apple-style-span" face="&#39;Segoe UI&#39;, Verdana, Arial"><a id="languageReferenceRemarksToggle"></a></font><p style="font-family: &#39;Segoe UI&#39;, Verdana, Arial; font-size: 13px; ">

This function is not atomic because it is implemented as two separate locked instructions. An atomic 64-bit read that occurs on another thread during the execution of this intrinsic could result in an inconsistent value being read.</p>

<div style="font-family: &#39;Segoe UI&#39;, Verdana, Arial; font-size: 13px; ">&quot;</div><div><meta charset="utf-8"><br></div><div style="font-family: &#39;Segoe UI&#39;, Verdana, Arial; font-size: 13px; "><br></div><div style="font-family: &#39;Segoe UI&#39;, Verdana, Arial; font-size: 13px; ">

<br></div></div><meta charset="utf-8">Can you claim, with absolute CERTAINTY that NOBODY in the WORLD depends on this behavior? That it&#39;s two separate intrusctions, and that an inconsistent value can be read? How do you know certain VMs don&#39;t look for this code sequence, how do you know security/emulation software doesn&#39;t abuse/utilise this? ETC ETC ETC. It&#39;s such a simple/short sequence that it&#39;s easy to recognize. </div>

<div><br></div><div>As a simple example, HexRays recognizes the particular ASM sequence as InterlockedAddLargeStatistic. Now, in ReactOS, it won&#39;t.</div><div><br></div><div>What&#39;s worse, is that recent Windows kernels define ExInterlocked.... to the compiler intrinsic _Interlocked. Do you suggest to also implement the intrinsic differently, which is in-lined? So now you will inline those 50 lines of (perfect, yes) code in every caller? These functions are called at ISR level usually -- great idea!</div>

<div><br></div><div>Or maybe you keep the (broken) MS implementation in the intrinsic, but use this one for the export? Great, now you introduce two sets of behaviors, and add even more non-determinism.</div><div><br></div>

<div>Do you really believe Microsoft is stupid and didn&#39;t realize this? That they didn&#39;t change things around just because they forgot? No, it&#39;s 20 years too late to change anything, and if you start doing this in ReactOS, you are certain to fail.</div>

<div><br></div><div>Please leave all personal attacks and politics aside, tell me where I&#39;m technically wrong. You cannot prove this better implementation won&#39;t break things. You have the compiler intrinsic to worry about. I named at least one piece of software that depends on it. I named critical speed/code size issues with this function. What more is needed to prove the point?</div>

<div><br></div><div>Best regards,<br>Alex Ionescu<br>
<br><br><div class="gmail_quote">On Sat, May 14, 2011 at 4:07 PM, Timo Kreuzer <span dir="ltr">&lt;<a href="mailto:timo.kreuzer@web.de">timo.kreuzer@web.de</a>&gt;</span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">

Am 15.05.2011 00:06, schrieb Alex Ionescu:<div class="im"><br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Congrats on once again breaking another key part of Windows compatibility.<br>
</blockquote></div>
Sorry for my ignorance. Could you please enlighten me and explain how code that behaves identically, except in very, very rare cases, where it doesnt fail miserably like the old version, and which might be a tiny piece slower, will break compatibility. Do drivers expect the function to fail randomly? Is the function time critical? Do snakes have legs? Whats the purpose of life?<div class="im">

<br>
<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
This function is DOCUMENTED to tear. It was DESIGNED that way on purpose.<br>
</blockquote></div>
I see, &quot;Its not a bug, its a feature&quot;(tm)<br>
<br>
<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
Utter failure, once more.<br>
</blockquote>
Of course, &quot;As usual&quot;(tm)<br>
<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
Best regards,<br>
Alex Ionescu<br>
</blockquote>
<br>
<br>
_______________________________________________<br>
Ros-dev mailing list<br>
<a href="mailto:Ros-dev@reactos.org" target="_blank">Ros-dev@reactos.org</a><br>
<a href="http://www.reactos.org/mailman/listinfo/ros-dev" target="_blank">http://www.reactos.org/mailman/listinfo/ros-dev</a><br>
</blockquote></div><br></div>