On Fri, 20 Dec 1996, Jason Thorpe wrote:

> It's smaller because it's linked against 4.4BSD libraries, which don't
> have YP code in them, etc.  It's mostly a library difference.
> 
> The MIT-SHM bit.. 4.4BSD had SYSV shared memory... don't know why that server
> doesn't use it... the comments in the file say it was for "HP-UX compatibility
> and X server (yech!)".  *shrug*
> 
> The reason for it not crashing... Well... heh, this is amusing, actually...
> 
> You suggest that some source should be imported into the NetBSD/hp300
> X server...
> 
> ...well, the amusing part is... it's the same source.  I.e. the sources
> used to build the 4.4BSD X server binary that works better are the _same_
> sources that were used to build the NetBSD/hp300 binary that occasionally
> loses.  (Well, a copy of them, anyhow...)
yep... I recently fetched the R5 sources and applied the patch that came
with the 4.4BSD X server... after some really minor work it compiled
happily, looked fast and - crashed. the heartbeat LED still flickered just
like normal, but a lot of moving "streaks" appeared on the screen and the
machine didn't do anything... for me it looked like a blitter error or
something, the streaks look like excessive video memory accesses ( at
least on some other video hardware memory accesses look like this ) and I
suspect that someone programmed the blitter to do an image height 0 or so
and the blitter counts down from 0 to 0... gives 2^32-1 operations which
should last a while and the rest of the machine waits patiently for the
blitter ( the heartbeat still goes... that's why I think of this... )
now... I think I will have a look at the lowlevel sources... does anyone
hav some documentation for the catseye-chip ? Or some suggestions ? I
find it really annoying...

anyway, I built the server without PEX, without MIT-SHM, and it still
crashes, so *WHAT* is really the difference between the 4.4BSD and the
NetBSD libraries / includes ? the compiler shoudn't be it, they used a
2.3ish gcc...

> Well, it's not a NetBSD binary... and, like you said, it lacks things
> like MIT-SHM.  The annoying bit is, it is probably impossible to reproduce
> that binary from sources anymore.
looks like this... doesn't somewhere a server exist which carries an old
utah-4.3BSD for the hp ?
 
> I would really much rather invest my, and have others invest their, effort
> into doing a decent X11R6 server for NetBSD/hp300.  We're destined to
> lose if we try to maintain an ancient 4.4BSD binary.
yup... what about the R6 server ? I thought about a simple bugfix in the
R5 but it seems to be more complicated...

bye
Michael