Building Emacs on SGI mips/netbsd

Building Emacs on SGI mips/netbsd

[Sascha Wilde]

Hi *,

I just build the current 23 head on an old SGI Indy (mips r4400) running
NetBSD 5.0:

To make it configure and build I had to fiddle with configure as it
currently does not handle NetBSD on mips and so the machine does not get
configured.

I suggest a patch along the lines of:

--8<---------------cut here---------------start------------->8---
--- a/configure.in      Mon Jun 15 14:21:29 2009 +0200
+++ b/configure.in      Wed Jun 17 14:24:12 2009 +0200
@@ -336,6 +336,10 @@
       x86_64-*-netbsd*)        machine=amdx86-64 ;;
       hppa-*-netbsd*)          machine=hp800 ;;
       m68k-*-netbsd*)          machine=m68k ;;
+      mips-*-netbsd* | mipseb-*-netbsd* \
+          | mips64-*-netbsd* | mips64eb-*-netbsd* )
+            machine=mips ;;
+
     esac
   ;;
 --8<---------------cut here---------------end--------------->8---

This is untested, as I actually manipulates the generated configure
script, and I'm not sure which exact targets are possible with netbsd
(IIRC it issued mipseb in my case, can't verify it at the moment).

Anyway, after hot-fixing configure it build and ran like a charm, so I
think once this is fixed we could add it to etc/MACHINES.

cheers
sascha
-- 
Sascha Wilde

"There is no reason why anyone would want a computer in their home"
Ken Olson, DEC, 1977

Re: Building Emacs on SGI mips/netbsd

[Glenn Morris]

For reference, this is bug #1609.

Re: Building Emacs on SGI mips/netbsd

[Stefan Monnier]

> +      mips-*-netbsd* | mipseb-*-netbsd* \
> +          | mips64-*-netbsd* | mips64eb-*-netbsd* )
> +            machine=mips ;;

Any reason why you didn't just use mips*-*-netbsd* ?


        Stefan

Re: Building Emacs on SGI mips/netbsd

[Miles Bader]

Stefan Monnier <monnier@iro.umontreal.ca> writes:
>> +      mips-*-netbsd* | mipseb-*-netbsd* \
>> +          | mips64-*-netbsd* | mips64eb-*-netbsd* )
>> +            machine=mips ;;
>
> Any reason why you didn't just use mips*-*-netbsd* ?

To avoid matching "mipshahanotreallyamips-*-netbsd"?

-miles

-- 
Dictionary, n.  A malevolent literary device for cramping the growth of
a language and making it hard and inelastic. This dictionary, however,
is a most useful work.

Re: Building Emacs on SGI mips/netbsd

[Glenn Morris]

Sascha Wilde wrote:

> +      mips-*-netbsd* | mipseb-*-netbsd* \
> +          | mips64-*-netbsd* | mips64eb-*-netbsd* )
> +            machine=mips ;;
> +

In Emacs 22, mips-*-netbsd*, mipsel-*-netbsd* and mipseb-*-netbsd*
used "machine=pmax". As far as I can tell, mips64* was not supported.

The file src/m/pmax.h has since been deleted. It does seem quite
important to me to fix this, but I could not say if mips.h is "good
enough" or if pmax.h should be restored.

Re: Building Emacs on SGI mips/netbsd

[Sascha Wilde]

Stefan Monnier <monnier@iro.umontreal.ca> wrote:

>> +      mips-*-netbsd* | mipseb-*-netbsd* \
>> +          | mips64-*-netbsd* | mips64eb-*-netbsd* )
>> +            machine=mips ;;

> Any reason why you didn't just use mips*-*-netbsd* ?

no other reason than having a quick look at some of the other rules in
configure.in -- actually as I wrote in my original post I'm not sure if
all these patterns are useful and/or needed.

cheers
sascha
-- 
Sascha Wilde

A conclusion is simply the place where someone got tired of thinking.

Re: Building Emacs on SGI mips/netbsd

[Sascha Wilde]

Glenn Morris <rgm@gnu.org> wrote:
> Sascha Wilde wrote:
>
>> +      mips-*-netbsd* | mipseb-*-netbsd* \
>> +          | mips64-*-netbsd* | mips64eb-*-netbsd* )
>> +            machine=mips ;;
>> +
>
> In Emacs 22, mips-*-netbsd*, mipsel-*-netbsd* and mipseb-*-netbsd*
> used "machine=pmax". As far as I can tell, mips64* was not supported.
>
> The file src/m/pmax.h has since been deleted. It does seem quite
> important to me to fix this, but I could not say if mips.h is "good
> enough" or if pmax.h should be restored.

Well, from my current experience I would say mips.h  is good enough at
least for non-mips64 R4000 systems (and afaik there is no mips64 netbsd at the
moment).  I don't have anything else to test, so people with different
hardware or more theoretical knowledge should probably verify this.  But
I'd suggest to re-enable mips/netbsd support asap, so that it can get some
wider testing.

cheers
sascha
-- 
Sascha Wilde
"Structure is _nothing_ if it is all you got.  Skeletons _spook_ people if
 thwy try to walk around on their own.  I really wonder why XML does
 not."            -- Erik Naggum <erik@naggum.net> in comp.lang.lisp

Re: Building Emacs on SGI mips/netbsd

[Glenn Morris]

Sascha Wilde wrote:

> Well, from my current experience I would say mips.h is good enough
> at least for non-mips64 R4000 systems (and afaik there is no mips64
> netbsd at the moment). I don't have anything else to test, so people
> with different hardware or more theoretical knowledge should
> probably verify this. But I'd suggest to re-enable mips/netbsd
> support asap, so that it can get some wider testing.

I have restored the mips netbsd support that was in Emacs 22.

Re: Building Emacs on SGI mips/netbsd

[Dan Nicolaescu]

Glenn Morris <rgm@gnu.org> writes:

  > Sascha Wilde wrote:
  > 
  > > Well, from my current experience I would say mips.h is good enough
  > > at least for non-mips64 R4000 systems (and afaik there is no mips64
  > > netbsd at the moment). I don't have anything else to test, so people
  > > with different hardware or more theoretical knowledge should
  > > probably verify this. But I'd suggest to re-enable mips/netbsd
  > > support asap, so that it can get some wider testing.
  > 
  > I have restored the mips netbsd support that was in Emacs 22.

The above seems to say that mips.h is good enough, why add again another
file for the same architecture (we already have 2 for mips: mips.h and
iris4d.h)?
Also, quite a bit of effort has been made to remove unused stuff in the
m/* files, is there anything interesting remaining in that pmax.h file
after removing #defines that are not used anymore?

Re: Building Emacs on SGI mips/netbsd

[Glenn Morris]

Dan Nicolaescu wrote:

> The above seems to say that mips.h is good enough, why add again another
> file for the same architecture (we already have 2 for mips: mips.h and
> iris4d.h)?
> Also, quite a bit of effort has been made to remove unused stuff in the
> m/* files, is there anything interesting remaining in that pmax.h file
> after removing #defines that are not used anymore?

I felt the conservative change of restoring the previous behaviour
most appropriate for the release branch. Feel free to change it on the
trunk.

Re: Building Emacs on SGI mips/netbsd

[Dan Nicolaescu]

Glenn Morris <rgm@gnu.org> writes:

  > Dan Nicolaescu wrote:
  > 
  > > The above seems to say that mips.h is good enough, why add again another
  > > file for the same architecture (we already have 2 for mips: mips.h and
  > > iris4d.h)?
  > > Also, quite a bit of effort has been made to remove unused stuff in the
  > > m/* files, is there anything interesting remaining in that pmax.h file
  > > after removing #defines that are not used anymore?
  > 
  > I felt the conservative change of restoring the previous behaviour
  > most appropriate for the release branch. Feel free to change it on the
  > trunk.

Thanks. I have done that on the trunk.