[prev in list] [next in list] [prev in thread] [next in thread] 

List:       openjdk-build-infra-dev
Subject:    sizer compilation has been broken for cross-compiles
From:       david.holmes () oracle ! com (David Holmes)
Date:       2012-11-27 3:27:47
Message-ID: 50B43333.6090205 () oracle ! com
[Download RAW message or body]

On 26/11/2012 11:51 PM, Fredrik ?hrstr?m wrote:
> 2012/11/26 David Holmes<david.holmes at oracle.com>:
>> If we can do this for cross-compiling why do we ever need to generate these files?
>
> If you read the makefile, you will see that it always uses the
> committed sources. The generation
> of the offset file is a verification step, run on platforms where it
> is possible, the generated offset file is not used per se.

Okay - I hadn't gleaned that from my quick scan of the change - sorry.

This needs to be blessed by the AWT folk of course, but it seems to me 
that if these values can change then it is not likely to be based on CPU 
architecture, so common/shared values should work. Otherwise this would 
all have to be based on something OS/runtime specific, in which case 
there would be no guarantee of a match between the end system and the 
build machine anyway.

Erik: just a follow up on one of your comments, the 32-bit/64-bit aspect 
of this was being handled by having the build host compiler do an 
appropriate build. At least that was how it worked in the old build and 
I'm pretty sure we made it do that in the new build too at some stage.

>> make the previous change? It still seems to me that things were working fine
>> before that.
>
> Obviously, there was a problem, that triggered the changed. In this case
> it was because of the latest MacOSX update that moved all the X11 libraries,
> so that you had to explicitly supply the path to them........

OSX continually causes problems. We must be assuming paths without 
realizing it if we keep needing to special case the OSX forms. :(

Thanks,
David


> //Fredrik

[prev in list] [next in list] [prev in thread] [next in thread]