[omniORB] Re: compiling v4.0.x with MSVC++ v5 sp3

[Dietmar May]

A known significant problem with MSVC++ 5.0 is that exceptions thrown 
from within a contructor are broken: the vtable-lookup code is off by 4 
bytes, and the attempt to throw an exception causes (in most cases) an 
access violation. The problem is fixed in 6.0.

I don't know if any of the omniORB code throws an exception from within 
a constructor -- if so, VC++ 5 will fail there sooner or later.

Dietmar

>>As I commented out the lines which caused the internal compiler errors,
>>eventually I got more mundane errors concerning socket structure field
>>names.  So my opinion at this point is that the combination of omniORB
>>v4.0.x and MSVC++ v5 sp3 would be very challenging to make work.  I'd
>>suggest removing all mention of MSVC++ v5 from the READMEs.
>>    
>>
>
>I don't have convenient access to VC++ 5, so I can't have a go
>myself. Can anyone confirm what Kendall says?  If so, and assuming
>nobody can come up with a sensible work-around, I'll update the list
>of supported platforms.
>
>Kendall (or anyone else), are you able to see if libraries compiled
>with VC++ 6 work with VC++ 5?
>