RFR: 8242244: Remove redundant ELF machine definitions

[Magnus Ihse Bursie]

On 2020-04-28 13:14, John Paul Adrian Glaubitz wrote:
> Hi Magnus!
>
> On 4/28/20 1:12 PM, Magnus Ihse Bursie wrote:
>> It does not work when I use the Oracle devkit. That linux/elf-em.h headers does not have the EM_AARCH64 define. Since this is a very actively used environment used to build OpenJDK, I'd say your assumption in the original bug "[...] unless the JDK is supposed to be built on systems older than that" was incorrect.
> Ok, thank you. Does the Oracle devkit define EM_486 or does that need to be there as well?
I thought a simple backing out would be the best, since we have a build 
breakage.

But sure, I understand why you want to keep part of the cleaning up.

I have checked, EM_486 is defined. So this combined partial backout + 
the fix suggested by Qi works:

diff --git a/src/hotspot/os/linux/os_linux.cpp 
b/src/hotspot/os/linux/os_linux.cpp
--- a/src/hotspot/os/linux/os_linux.cpp
+++ b/src/hotspot/os/linux/os_linux.cpp
@@ -110,6 +110,7 @@
  # include <stdint.h>
  # include <inttypes.h>
  # include <sys/ioctl.h>
+# include <linux/elf-em.h>

  #ifndef _GNU_SOURCE
    #define _GNU_SOURCE
@@ -1845,6 +1846,9 @@
      char*         name;         // String representation
    } arch_t;

+#ifndef EM_AARCH64
+  #define EM_AARCH64    183               /* ARM AARCH64 */
+#endif
  #ifndef EM_RISCV
    #define EM_RISCV      243               /* RISC-V */
  #endif


I'd say that this is a trivial fix, and you can consider it reviewed by 
me (although the authorship of this patch is somewhat convoluted).

/Magnus

>
>> Recommending that you back out this change.
> Apologies for the breakage. I should have done more testing.
>
> Adrian
>