Small Request for review: 7005007 - Refine use of ALT_COMPILER_PATH to avoid conflict with JPRT usage

David Holmes David.Holmes at oracle.com
Tue Dec 7 13:54:05 PST 2010


Kelly,

I can file a RFE but repairing the build system is out of scope for this 
particular CR.

David

Kelly O'Hair said the following on 12/08/10 02:36:
> 
> On Dec 7, 2010, at 3:26 AM, David Holmes wrote:
> 
>> Simple fix to avoid conflicting use of ALT_COMPILER_PATH.
>>
>> http://cr.openjdk.java.net/~dholmes/7005007/
>>
>> Tested with JPRT linux builds and doing an internal cross-compile build.
>>
>> Thanks,
>> David Holmes
> 
> IF nm doesn't exist, then this shell script will fail, but return exit 
> code 0, and the makefile
> will not stop but create a bogus mapfile, which may or may not allow 
> libjvm.so to link.
> I think this shell script needs to use 'set -e' or do something like (nm 
> || exit 1) | awk
> or maybe both.  Can we fix this shell script?
> 
> In addition. I'd like to see some makefile verifications that the tools 
> actually exist, something like
> 
> # Check fullpath tool
> define filecheck # fullpath
> $(if $(wildcard $1),$1,$(error "ERROR: File does not exist: $1"))
> endef
> 
> # Check tools
> ifdef CROSS_COMPILE_ARCH
>   ifdef ALT_COMPILER_PATH
>     CC      := $(call filecheck,$(ALT_COMPILER_PATH)/gcc)
>     CPP    := $(call filecheck,$(ALT_COMPILER_PATH)/g++)
>     NM     := $(call filecheck,$(ALT_COMPILER_PATH)/nm)
>     STRIP := $(call filecheck,$(ALT_COMPILER_PATH)/strip)
>   endif
> endif
> 
> -----
> We also need to make sure we document this in the README-builds.html 
> file, but that can be
> done some other time since it is in a different repository..
> 
> -kto
> 


More information about the hotspot-runtime-dev mailing list