Repository navigation
Update z/OS compile to use V2R3 - #6250
Conversation
|
I am thinking we need to differentiate JDK8 and JDK11 if we have different support levels. JDK8 = V2R1 iirc, whereas JDK11 = V2R3. Do I have this correct? Something like how OpenJ9 already has it? Example |
|
OpenJ9 overrides OMR settings so that would appear irrelevant to differentiate for JDK8/11. Do we want to match the system version of V2R4 for OMR_ZOS_COMPILE_TARGET and OMR_ZOS_LINK_COMPAT? OMR_ZOS_COMPILE_ARCHITECTURE...@joransiu / @r30shah I am thinking this is linked to the version of xlc or system version??. I see openj9 has this value defaulting to 10 with V2R3? What where should this value be? |
| @@ -114,9 +114,9 @@ elseif(OMR_OS_LINUX) | |||
| ) | |||
| elseif(OMR_OS_ZOS) | |||
| set(OMR_ZOS_COMPILE_ARCHITECTURE "7" CACHE STRING "z/OS compile machine architecture") | |||
There was a problem hiding this comment.
If we go ahead with this change, this should change to "10" to match the V2R3 settings.
| @@ -1747,11 +1747,11 @@ Some influential environment variables: | |||
| OMR_ZOS_COMPILE_ARCHITECTURE | |||
| The z/OS compile machine architecture. (Default: 7) | |||
| @@ -2880,9 +2880,9 @@ OMR_PRODUCT_NAME=${OMR_PRODUCT_NAME:-"OMR"} | |||
| OMR_PRODUCT_VERSION=${OMR_PRODUCT_VERSION:-"0.1"} | |||
|
|
|||
| OMR_ZOS_COMPILE_ARCHITECTURE=${OMR_ZOS_COMPILE_ARCHITECTURE:-"7"} | |||
| @@ -69,11 +69,11 @@ AC_ARG_VAR([OMR_PRODUCT_DESCRIPTION], | |||
| AC_ARG_VAR([OMR_ZOS_COMPILE_ARCHITECTURE], | |||
| [The z/OS compile machine architecture. (Default: 7)]) | |||
@jdekonin , Java 11 on z/OS supports V2R3 and later on zEC12, that is why we have |
|
This change may break Java 8 builds, as not all places set the desired value for 8, but rely on the OMR defaults. |
Java 11 on z/OS actually pre-reqs V2R4, though we are still compiling on V2R3 internally. That said, as noted above already, Java 8 will still need to be built with the original V1R13 compatibility. |
|
The reason I did this change is because on the new V2R4 guests the use of V1R13 produced warnings like I have detailed below. OMR doesn't appear to have a notion of building for a specific JDK release, so it would appear the options are
I'll test with what the PR currently has to see what the the breakage is like for Java 8 builds. |
|
We can modify OpenJ9 if necessary to specify the correct values when building Java 8, that just needs to be done first. |
|
It would appear from my internal testing that this change is in fact completely overwritten by openj9. Being that our original 2 z/OS guests are on V2R2, this will need to wait until 1 or both new systems are online. |
|
I had to push a new change for the automated build. The user id and home directory is different on the new systems, so this change should make that more flexible for any future machine adds that are not identical. |
|
Jenkins build zos_390-64 |
|
Jenkins build zos |
|
Jenkins build all |
|
Jenkins build zos_390-64 |
1 similar comment
|
Jenkins build zos_390-64 |
|
Seems PR jobs don't use the PR code change in the job when it comes to the build scripts. |
|
Still needs to hold until the 2nd z/OS guest from Marist is availlable. eta today/tomorrow I hope |
|
Jenkins build all |
6e54924 to
021926f
Compare
Issue eclipse-omr/omr#6250 (comment) Signed-off-by: Peter Shipton <Peter_Shipton@ca.ibm.com>
|
Performed a rebase and new round of builds to verify. The older systems have been requested to be returned so unless there are additional changes needed this requires review/merge. Discussed with @pshipton & @AdamBrousseau via slack and there should be no dependency or problems with OpenJ9. |
|
Jenkins build zos |
2 similar comments
|
Jenkins build zos |
|
Jenkins build zos |
* [skip ci] * modify hardcoded user home to variable Signed-off-by: Joe deKoning <joe_dekoning@ca.ibm.com>
Signed-off-by: Joe deKoning <joe_dekoning@ca.ibm.com>
|
Jenkins build all |
|
z/OS job in the build all was directed to |
|
Jenkins build zos |
|
@joransiu : can you please formally approve this if you think it is ready to merge? |
They need to be explicitly set for Java 8 because the defaults were changed by eclipse-omr/omr#6250 Signed-off-by: Peter Shipton <Peter_Shipton@ca.ibm.com>
Signed-off-by: Joe deKoning joe_dekoning@ca.ibm.com