[sldev] Call stack for crash on startup in SSE code on early Athlon processors?

[Dzonatas]

I have just borrowed more hardware to further test this issue. The exact 
result will not be release, however. That is due to an equivocal 
agreement with Zero Linden made in the office hours under the envelope 
of Quality Assurance and the Residents. We might see some progress for 
LL to extend the "team" effort of required by QA to *actually* involve 
all developers that have worked on Second Life.

I'm going to put a higher priority on the issue to get the 
auto-initialized global C++ classes moved over to be manually 
initialized. For one, it is obviously impossible to fully QA such 
classes with them being auto-initialized in a chaotic order. There are 
just dozens of classes that could potentially output very useful log 
information, but such classes can't until the log class is initialized 
and opens the log file, which currently happens after the 
auto-initialized classes!

You know I have involuntarily wasted time to trace auto-initialized C++ 
classes when it could have been done much much easier!

Dirk Moerenhout wrote:
> Which binary did you test? Given this is more a compiler issue than a
> code issue I doubt it'll crash if you were using anything but an MSVC
> compiled binary. And even then it might be compiler version dependent.
>
> Dirk aka Blakar Ogre
>
> On 7/10/07, Dzonatas <dzonatas at dzonux.net> wrote:
>>
>>  The immediate test on the 586tsc shows it working -- starts up. The 
>> 586tsc
>> does not have SSE but does have MMX. The only thing apparently different
>> with the K7 would be the 3DNow! features.
>>
>>  Dzonatas wrote:
>>  I got a hold of a 586tsc, which appears very similar to the older
>> Thunderbird 800mhz.  Eh... it has linux preinstalled... i'll see what 
>> i can
>> do
>>
>>  Harold Brown wrote:
>>
>> Not sure what I all I can do with the system I tested with last night.
>> (Athalon XP 1500+, No SSE/SSE2)
>>
>> It's not exactly the beefiest machine but I'll see what I can do.  
>> I'm off
>> RL work for vacation starting this weekend.. so I'll definately have 
>> some
>> time to mess with it then.
>>
>> On 7/10/07, Dzonatas <dzonatas at dzonux.net> wrote:
>>
>> > James, let me know and I'll drive out there to help you set it up and
>> > work on it... =)
>> >
>> > James Cook (James) wrote:
>> > > I tried and failed yesterday to fix "Crash on startup due to SSE
>> > > instructions":
>> > > https://jira.secondlife.com/browse/VWR-1610
>> > >
>> > > We mistakenly tested this build on an Athlon Mobile XP 2000+ which
>> > > supports SSE1.
>> > >
>> > > None of our internal developers have a processor this old.  Does
>> > > anyone have an early Athlon (K7 / Thunderbird) that doesn't support
>> > > SSE who can get me a call stack off the current codebase?
>> > >
>> > > I'm reverting the SSE code from 1.18.0 until we can sort this out.
>> > > We're putting together an old Athlon internally, but it won't be in
>> > > time for the 1.18 series.
>> > >
>>
>>
>> -- 
>>  Power to Change the Void ________________________________
>>
>> _______________________________________________
>> Click here to unsubscribe or manage your list subscription:
>> /index.html
>>
>>
>>
>> -- 
>>  Power to Change the Void
>> _______________________________________________
>> Click here to unsubscribe or manage your list subscription:
>> /index.html
>>
>>
>
>

-- 
Power to Change the Void