Jonathan Stone wrote:
> 
> 
> >I just rebuild egcs and with that a kernel, all from sources supped this
> >morning.
> >
> >The we driver still panics on sending the first ethernet packet.
> >
> >Please tell me if I can do anything to help hunting this down...
> 
> short of a more targeted suspicion, the first step is to try
> recompiling
>  suspect kernel .o files  with the old compiler, to see where the problem
> goes away -- to find the footprint of this source/compiler bug.
> 
> I'd try recompiling all the if_we related code: dp8390, whichever
> bus-specific front-end attachment (ISA, PCI -- only ISA for the we,
> right??) you're using, and possibly all the mbuf code as well.
> That's:
> 
> 	dp8390.o
> 	if_we.o
> 	uipc_mbuf.o
> 
> for starters. If recompiling them with the old complier makes your
> kernel work, then you've helped (and ideally keep going, pin it down
> to one file).  if it doesn't, expand the search.
> 
> 
> Todd: are ever we going to wake up to the fact that initial versions
> of compilers *are* usually buggy (or uncover source-code bugs), on at
> least some of our arches?  Why don't we provide intree support for
> keeping the previous version of the compiler binaries and specfiles
> around for cases like this?  It's not like gcc/egcs doesnt already
> ship with all the prefixing necessary to do this...
> 

I now doubt that my i386 woes are to do with the compiler, but rather
with the kernel, though I didn't try with gcc-2.7.2. Kernel compiled
with
1) egcs 1.1.1 compiled by me
2) egcs 1.1.1 from snapshot
3) egcs 1.1 from a working box, compiled 1 October
all result in a system where the respective (different) compiler
receives signal 11s. Could one of you send me i386 kernel source from
say 14th November?

Cheers,

Patrick