[M3devel] more info on juno on windows

[Tony Hosking]

PS  Running with nogc at least makes sure that you don't get the  
garbage collector moving things around, so it should be easier to  
diagnose who is doing the clobbering.  Can you debug with h/w  
watchpoints to see who overwrites the heap.  You'd need to watch the  
location from which the bogus reference (00200000) gets loaded, which  
means figuring out where the load occurred.

On 22 Sep 2009, at 17:58, Randy Coleburn wrote:

> Tony:
>
> I just tried these options.  Here are results:
>
> recent cm3 on XP:  paranoidgc yeilds assert in RTCollector.m3 line  
> 1706, while nogc gets assert in WinContext.m3 line 165.  I note that  
> the juno window begins drawing before the crash on nogc whereas it  
> does not on paranoidgc.
>
> recent cm3 on Vista, same results as above except that it appears to  
> reference an illegal memory location before hitting the assert in  
> the RTCollector when using paranoidgc.
>
> old d5.7.0 circa August 2008 on XP:  paranoidgc gets a repeating  
> assert at line 845 in ThreadWin32.m3.  You have to hit Ctrl-C to  
> abort the repeating error message.  Not sure if anything else  
> happens first because it scrolls too far.  For nogc, we get same  
> behavoir as the other tests above.
>
> Regards,
> Randy
>
> >>> Tony Hosking <hosking at cs.purdue.edu> 9/22/2009 5:46 PM >>>
> Have you tried running with @M3nogc?  And @M3paranoidgc?
>
>
> Antony Hosking | Associate Professor | Computer Science | Purdue  
> University
> 305 N. University Street | West Lafayette | IN 47907 | USA
> Office +1 765 494 6001 | Mobile +1 765 427 5484
>
>
>
>
> On 22 Sep 2009, at 17:39, Jay K wrote:
>
>>
>> Again, what I see is that many versions before around Feb 20 2007  
>> consistently fail with that same assertion failure.
>>
>> I have tested many versions now, recently.
>>
>> But versions after Feb 20 2007 usually access violate on the  
>> address 0x20000 or so, sometimes other addresses, sometimes various  
>> assertion failures. I believe this is much worse than merely always  
>> failing the same assertion.
>>
>>
>>
>> - Jay
>>
>>
>>
>> Date: Tue, 22 Sep 2009 17:06:20 -0400
>> From: rcoleburn at scires.com
>> To: m3devel at elegosoft.com
>> Subject: [M3devel] more info on juno on windows
>>
>>
>>
>>
>> Do we know whether or not Juno ever worked on Windows ?
>>
>> I don't recall ever testing it on Windows.  I still have a vd5.7.0  
>> cm3 that I used for the project I finished up last year (August  
>> 2008).  If I run Juno on this system (Windows XP SP3), Juno crashes  
>> with an ASSERT failure at line 165 in winvbt/WinContext.m3.  The  
>> date on the juno.exe is 8/19/2008.
>>
>> Regards,
>> Randy
>>
>>>>> Jay K <jay.krell at cornell.edu> 9/22/2009 2:57 PM >>>
>> Here is the truncated part from the previous:
>>
>> This change, I think, causes Juno to access violate whereas before  
>> it "only" failed assertions.
>> I believe it is considered fairly ok for a safe system to terminate  
>> with an assertion failure,
>> that might not be a bug at all, but considered far worse to hit a  
>> SIGSEGV
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://m3lists.elegosoft.com/pipermail/m3devel/attachments/20090922/3f711d62/attachment-0002.html>