Re: [Qemu-devel] [patch 1/2] machine struct - use C99 initializers

[Paul Brook <paul@codesourcery.com>]

> : If a host system hasn't bothered upgrading their toolchain in 10 years
> : then I refuse to care. If you really want to run and ancient obsolete OS
> : you should expect to run equally ancient software.
>
> You assume that all upgrades are a good thing.  There are often
> serious regressions in newer software, especially in not Intel
> platforms, that makes it much harder to upgrade and have a working
> system afterwards.
>
> I'm saying there needs to be a balance between the latest and
> greatest, and known working software...

IMHO anything using gcc2 has the balance tipped completely the wrong way, and 
requires a disproportionately large amount of effort to support.  I make a 
distinction between systems that are a few of revisions behind the bleeding 
edge, and systems that are not making enough effort to stay current. If we 
were requiring gcc4 then I'd agree with you. However I think sufficient time 
has passed that it's reasonable to expect at least gcc3.

As a side note, I'd be surprised if most modern c++ code even compiled with 
gcc2.95. The language implemented by g++2.95 only bears passing resemblance 
to modern c++, which is the reason the ABI is completely different.

> [Jes wrote]
> One interim step could be for OSes like BeOS to switch to egcs or Red
> Hat's old gcc-2.96 base. I believe both support C99 style struct
> initializers while still sticking to the old C++ ABI.

I wouldn't touch "2.96" unless you absolutely have to (i.e. you're still 
running RH7 systems). There's about a million different variants, most of 
which are broken.

Paul