RFR: JDK-8031759 Configure should handle overrides of tools better

Magnus Ihse Bursie magnus.ihse.bursie at oracle.com
Wed Jan 22 15:29:35 UTC 2014


Sure, I forgot about the greadlink addition. I'll fix that. 

Also, your comment got me wondering if this is the wrong solution after all. We have tried hard to stay clear of letting environment variables affect the build, since they are "invisible", impossible to check and creates hard to debug problems. And now I'm letting in a hoard. :-/ Overriding tools is one thing, but I'd really prefer to do it on the configure command line. 

I'll see if I can tweak configure to only accept overrides as command line arguments, and emit warnings or errors if an incorrect override is attempted. 

Stay tuned for a new webrev. 

/Magnus

> On 22 jan 2014, at 02:29, Mike Duigou <mike.duigou at oracle.com> wrote:
> 
> The change works for me and does meet the needs I was trying to address in 8031669.
> 
> Could you include the addition of [greadlink readlink] that was in my patch? The Mac version of this tool from either fink or macports is named greadlink and has advantages over the manual resolution method (most directly it understands case insensitive file systems).
> 
> I am also wondering if many of the tools we rely upon and whose location can be set via environment vars should be documented via AC_ARG_VAR. The rationale being that if we are pulling some element of the config from a var then we should be documenting that the var is significant.
> 
> Mike
> 
>> On Jan 20 2014, at 06:57 , Magnus Ihse Bursie <magnus.ihse.bursie at oracle.com> wrote:
>> 
>> Bug: https://bugs.openjdk.java.net/browse/JDK-8031759
>> 
>> In a way, this is a continuation of JDK-8022411. While that bug fixed the ability to properly override CC/CXX with either a complete path, or just a command name that was to be searched in $PATH, this fix does the same for all other tools that are located, not only the compilers. (Boot-JDK tools being the exception; this is issue might be addressed soon too.)
>> 
>> With this fix, user overrides will always be honored, and they will never fail silently.
>> 
>> I believe this fix supersedes the pending fix for JDK-8031669, which is still awaiting approval from a formal reviewer.
>> 
>> WebRev: http://cr.openjdk.java.net/~ihse/JDK-8031759-better-overriding-of-tools-in-configure/webrev.01
>> 
>> /Magnus
> 



More information about the build-dev mailing list