Is "host" the build or the target system?

[Kelly O'Hair]

On May 24, 2012, at 2:13 PM, Fredrik Öhrström wrote:

> Hey guys! You are falling into the old tar pit of "not invented here" mentality that Sun was rather (in)famous for. 

I politely disagree, and am a little insulted, being a long term Sun developer.

Innovation comes about by changing things. Nobody has a monopoly on innovation.

> 
> If there is a standard, then we should follow it.

And standards change, and many standards die because people don't want to use them because the standardization
process forgot about getting consensus from the community.

I'll be willing to bet that this autoconf change had quite a few complaints when it was rolled out.

> 
> Let me quote from the GNU autoconf manual
> "Based on the experience of compiler writers, and after long public debates, many aspects of the cross-compilation chain have changed:" and "The relationship between build, host, and target have been cleaned up:"
> 
> Thus, the people responsible for autoconf/configure have already rehashed this several times. This is what they came up with,
> and this is what is described on all pages when you search for cross compile, and this is what is described on wikipedia about
> cross compiling and autoconf, and this is what is used by other projects that use configure and autoconf.

Doesn't make it right in my view. I think they messed up.

But this is all water under the bridge.

> 
> We cannot just change the default behavior of configure, just because it feels better with some other english words....

I accept your position that we should try to not change expected autoconf behavior here if we can avoid it.

I myself am ok with just cringing everytime I use the --host options, but I still consider it an autoconf BUG,
always will. 

> 
> You are running on i686-unknown-linux-gnu, 
> you want to build a cross compiler, then use:
> --target=sparc-sun-solaris2_10 when configuring gcc
> 
> Now you want to use your newly created cross compiler (sparc-sun-solaris2_10-gcc in your path)
> to build a binary of application foo, for your sparc:
> --host=sparc-sun-solaris2_10 when configuring foo

I bet I could count on one hand the number of people that would actually want to build a cross compiler
for the JDK, this is such a rare event.

> 
> To hack configure to alias --target to --host will not make life easier for us and it will introduce
> a use case that we will have to support for ever, and this might be difficult in the future when
> autoconf is upgraded.

You do have a point on the upgrade.

> 
> At the moment, we cannot compile static binaries from the OpenJDK, we only run within a runtime,
> but it clearly is not farfetched to assume that such a feature would be desirable. gcj has it already.
> Thus --target might be valuable for exactly its specified purpose some time in the future.

Yeah right. ;^)  Is LSD legal in Sweden? ;^)

> 
> Sorry for the quick backout, but I had to back out the change quickly since it affected quite a lot of variables
> and it would be inconvenient if someone managed to do another merge after it.

You do what you need to do.

----

But seriously, you are making an argument that we should not try and modify the meaning or behavior
of the autoconf options because it will only serve to confuse matters in the future and for existing autoconf users.
So even though I think it's autoconf that has created the confusion, I see your point.

If we accept that we are adopting autoconf, which I think we did, then we probably shouldn't try and change it's
basics and stick with just adding our special options.

I'm not sure I have quite the objection to the compromise you had, but I will defer to your autoconf expertise
in this matter. Let's see if you convince Magnus. ;^)

-kto

> 
> //Fredrik
> 
> ----- kelly.ohair at oracle.com skrev:
> 
>> Ok.  You've sold me the bridge. I may need some time and a few glasses
>> of wine, but I think I am with you.
>> 
>> I'm still a little shocked at the autoconf terminology.  Seems like
>> they completely complicated things for a situation
>> that was extremely rare and few people would ever need to do or even
>> care about.
>> 
>> For the record, I'm in favor of your compromise.  Thanks for the time
>> in explaining it to this autoconf novice, me. ;^)
>> 
>> -kto
>> 
>> On May 24, 2012, at 1:30 PM, Magnus Ihse Bursie wrote:
>> 
>>> On 2012-05-24 17:00, Kelly O'Hair wrote:
>>>> So we should talk about this, and maybe come up with a compromise
>> or alternative terminology.
>>> 
>>> Just to be clear: My suggestion was to use "build system" for the
>> platform you are building on, and "target system" for the platform the
>> resulting binary should run on, and that we should stay clear of
>> "host".
>>> 
>>> Rationale: "build system" and "target system" seems hard to
>> misunderstand, regardless of your previous experiences.
>>> 
>>> "Host" on the other hand, is vague enough to be able to be used in
>> the sense of "build system" (as in the old openjdk makefiles) or as
>> "target system" (as in the autoconf tradition), depending (I guess) if
>> your mindset is centered on the build process or the resulting binary.
>> So I say: don't use it at all.
>>> 
>>> This means that:
>>> * The openjdk tradition of calling the target system "target" is
>> kept, but the build system will be renamed from "host" to "build".
>>> * The autoconf tradition of calling the build system "build" is
>> kept, but the target system will be renamed from "host" to "target".
>>> 
>>> I think it sounds like a good compromise. :)
>>> 
>>> /Magnus