Changes to be pushed into jdk8/build

[Magnus Ihse Bursie]

On 2012-09-13 02:09, David Holmes wrote:
> - common/autoconf/configure.ac
>
> Minor typo:
> + # Make the comaper script executable
I've already submitted a fix.

>
> ---
>
> - common/autoconf/platform.m4
>
> 408 # Make compilation sanity check
>  409 AC_CHECK_HEADERS([stdio.h], , [
>  410   AC_MSG_NOTICE([Failed to compile stdio.h. This likely implies 
> missing compile dependencies.])
>  411   AC_MSG_NOTICE([If you are using reduced build, try installing 
> package for libc6-dev-i386 or similar])
>  412   AC_MSG_ERROR([Cannot continue.])
>  413 ])
>
> Not sure if this will "just work" on full cross-compiles. Also the 
> message seems too linux specific (and you need lots of i386 packages 
> installed to do full 32-bit builds on 64-bit linux).

I agree that the message is bad. My fault, I started it out with the 
ambition of getting --with-target-bits=32 to work on my Linux x86_64 
machine, and to at list at least what Ubuntu packages was needed. Then 
it ended up being a general compilation sanity check, and I never 
completed the list of packages anyway. :-)

However, I do believe it will "just work" on cross-compiles. What it 
does is to create a simple program, like test.c:
#include <stdio.h>
and then compile it using the toolchain we're about to set up. If this 
doesn't work, then you have a lot of trouble, and basically nothing else 
is ever going to compile. I wanted to catch that case quickly, otherwise 
the output from configure gets really messed up and confusing.

I'll change the message to something more neutral.


> ---
>
> jdk/src/share/demo/jvmti/hprof/hprof_init.c
>
> Not sure what this change has to do with anything :) Seems like the 
> code involved here is pretty confused about data type sizes.
Yeah. I wonder who changed this and why.

There are several other hotspot changes like this one that are quite 
unrelated to the rest of the changes, and that we probably should not 
push at this time (if at all).

/Magnus