Re: DEC PWS433a and 2.1.61VGER don't match...

Matti Aarnio (matti.aarnio@tele.fi)
Wed, 12 Nov 1997 01:23:09 +0200 (EET)


Richard Henderson comments on my email:
> > Most of what appears in the source with (vuip) cast are macroes,
> > and it may well be that a macro-expansion surprise has bitten
> > in here... Hmm.. No, can't say it definitely, but most of
> > include/asm-alpha/pyxis.h defined values are within parens.
>
> Indeed, I thought I'd made sure of that when I made the change. I
> don't suppose you could save me a bit of trouble and send me the
> preprocessed pyxis.c files for the working and non-working kernel?
> I'd like to see what difference is made in the assembly.

I did compile unmodified and modified pyxis.c into pyxis.s
files (same command line for both runs).

Doing a diff revealed NO difference :-(
The gcc and pyxis.c have thus been releaved
from that suspicion.

Which means my next reboot may not be as smooth as
the previous one was (if something like 9 out of 10 fails
as the previous rate has been...)

> > (I have no idea, why same source on intel box did
> > not need special tricks for producing working ipv6 module.)
>
> The intel ipv6 csum magic routine is inline, isn't it?

Propably. (Those who don't know, Richard made separate
assembler module for that routine, which requires special
linkage to be usable in modular IPv6.)

> r~

/Matti Aarnio