Comments on: Delphi bug of the day: FPU stack leak http://blog.excastle.com/2007/08/28/delphi-bug-of-the-day-fpu-stack-leak/ Life, .NET, and Cats Sun, 12 Oct 2008 07:33:51 +0000 http://wordpress.org/?v=2.5 By: Ritchie Annand http://blog.excastle.com/2007/08/28/delphi-bug-of-the-day-fpu-stack-leak/#comment-6925 Ritchie Annand Fri, 21 Mar 2008 06:32:47 +0000 http://blog.excastle.com/2007/08/28/delphi-bug-of-the-day-fpu-stack-leak/#comment-6925 A little more detail (but not much) on the Cisco trouble: it's on reallocating strings, and basically it seems to be munching the very first fild - either popping it off the FPU stack or resetting it, because you end up with an *underflow*, and if you mask out the invalid floating point operation, you end up with the first bytes (which are the first grabbed and put on the FPU stack) garbled. It would be nice if OllyDbg (a very nice tool for most assembly tracing) could log out more of the FPU registers (it just seems to do the ST(0) through ST(7)) so I could trace this thing, but alas... A little more detail (but not much) on the Cisco trouble: it’s on reallocating strings, and basically it seems to be munching the very first fild - either popping it off the FPU stack or resetting it, because you end up with an *underflow*, and if you mask out the invalid floating point operation, you end up with the first bytes (which are the first grabbed and put on the FPU stack) garbled.

It would be nice if OllyDbg (a very nice tool for most assembly tracing) could log out more of the FPU registers (it just seems to do the ST(0) through ST(7)) so I could trace this thing, but alas…

]]> By: Ritchie Annand http://blog.excastle.com/2007/08/28/delphi-bug-of-the-day-fpu-stack-leak/#comment-5823 Ritchie Annand Mon, 11 Feb 2008 19:07:26 +0000 http://blog.excastle.com/2007/08/28/delphi-bug-of-the-day-fpu-stack-leak/#comment-5823 Thanks for posting this. We've been running into some very strange Invalid Floating Point Operation errors ourselves. In this case, though, the cause is not internal to the program, but rather, from Cisco VPN v4.6 on some machines. Basically, we do socket reads, and at random intervals (I saw something in the packet trace log that looked like a broken packet it was trying to resend, but I can't be sure that was the issue) it sets the FPU into a state that is ripe for causing the System.Move to have an invalid floating point operation. It's not limited to Delphi programs by any means. Matlab (in particular!) and SQL Server Merge Replication are also sensitive to the issues in Cisco's v4.6. Thanks for posting this.

We’ve been running into some very strange Invalid Floating Point Operation errors ourselves. In this case, though, the cause is not internal to the program, but rather, from Cisco VPN v4.6 on some machines. Basically, we do socket reads, and at random intervals (I saw something in the packet trace log that looked like a broken packet it was trying to resend, but I can’t be sure that was the issue) it sets the FPU into a state that is ripe for causing the System.Move to have an invalid floating point operation.

It’s not limited to Delphi programs by any means. Matlab (in particular!) and SQL Server Merge Replication are also sensitive to the issues in Cisco’s v4.6.

]]> By: A26gen http://blog.excastle.com/2007/08/28/delphi-bug-of-the-day-fpu-stack-leak/#comment-5690 A26gen Fri, 21 Dec 2007 07:58:00 +0000 http://blog.excastle.com/2007/08/28/delphi-bug-of-the-day-fpu-stack-leak/#comment-5690 Hi, <br> <br>Thanks for this great post ! <br> <br>I wanted to try the repro test but it seem's that the report doesn't exist anymore on QualityCentral. Is it possible to post it again and to give us the new link please. <br> Hi,

Thanks for this great post !

I wanted to try the repro test but it seem’s that the report doesn’t exist anymore on QualityCentral. Is it possible to post it again and to give us the new link please.

]]> By: Birger http://blog.excastle.com/2007/08/28/delphi-bug-of-the-day-fpu-stack-leak/#comment-5689 Birger Wed, 19 Dec 2007 12:55:00 +0000 http://blog.excastle.com/2007/08/28/delphi-bug-of-the-day-fpu-stack-leak/#comment-5689 Thanks for this blog. I came across it when searching for floating point and stack errors and it helped us explaining a problem we encountered! Thanks for this blog. I came across it when searching for floating point and stack errors and it helped us explaining a problem we encountered!

]]> By: Rick http://blog.excastle.com/2007/08/28/delphi-bug-of-the-day-fpu-stack-leak/#comment-5688 Rick Thu, 30 Aug 2007 22:43:00 +0000 http://blog.excastle.com/2007/08/28/delphi-bug-of-the-day-fpu-stack-leak/#comment-5688 Thanks for taking the time to BLOG about this in any easy to read manner. It's been many years since I coded in ASM and I somtimes forget what the compiler does for us when using higher level languages like Pascal. <br> <br>This also begins to explain FastMM to provide such an improvement... something I've always curious about but not important enough to investigate. <br> <br> Thanks for taking the time to BLOG about this in any easy to read manner. It’s been many years since I coded in ASM and I somtimes forget what the compiler does for us when using higher level languages like Pascal.

This also begins to explain FastMM to provide such an improvement… something I’ve always curious about but not important enough to investigate.

]]> By: william http://blog.excastle.com/2007/08/28/delphi-bug-of-the-day-fpu-stack-leak/#comment-5687 william Wed, 29 Aug 2007 23:51:00 +0000 http://blog.excastle.com/2007/08/28/delphi-bug-of-the-day-fpu-stack-leak/#comment-5687 Nice read with detailed explanation! Nice read with detailed explanation!

]]> By: David M http://blog.excastle.com/2007/08/28/delphi-bug-of-the-day-fpu-stack-leak/#comment-5686 David M Wed, 29 Aug 2007 23:02:00 +0000 http://blog.excastle.com/2007/08/28/delphi-bug-of-the-day-fpu-stack-leak/#comment-5686 Interesting read! Interesting read!

]]> By: Craig Stuntz http://blog.excastle.com/2007/08/28/delphi-bug-of-the-day-fpu-stack-leak/#comment-5685 Craig Stuntz Wed, 29 Aug 2007 12:05:00 +0000 http://blog.excastle.com/2007/08/28/delphi-bug-of-the-day-fpu-stack-leak/#comment-5685 Thanks; I reproduced this and opened the report. Thanks; I reproduced this and opened the report.

]]>