arm soft/hard float detection

[croudet]

• jna v4.5.1
• armv7l 32
• ubuntu based linux (nvidia tk1)
• jdk-8u162-linux-arm32-vfp-hflt
  readelf -h ./jdk1.8.0_162/bin/java
  ELF Header:
  Magic: 7f 45 4c 46 01 01 01 00 00 00 00 00 00 00 00 00
  Class: ELF32
  Data: 2's complement, little endian
  Version: 1 (current)
  OS/ABI: UNIX - System V
  ABI Version: 0
  Type: EXEC (Executable file)
  Machine: ARM
  Version: 0x1
  Entry point address: 0x8484
  Start of program headers: 52 (bytes into file)
  Start of section headers: 2388 (bytes into file)
  Flags: 0x5000002, has entry point, Version5 EABI
  Size of this header: 52 (bytes)
  Size of program headers: 32 (bytes)
  Number of program headers: 8
  Size of section headers: 40 (bytes)
  Number of section headers: 31
  Section header string table index: 28
• related to Investigate supporting arm soft- and hardfloat #753

Arm soft/hard float detection does not work because the java exe (ELF header) does not contain the required flags.
In version 4.5.0 ElfAnalyser returned false for both soft and hard float and by luck, the hard float lib was chosen.
In 4.5.1 the detection has slightly changed and now the soft float flag is set to true.

I wonder if a system property can be added (e.g: jna.arm.hardfloat=true) to set the hard/soft flag and to disable runtime detection.

Thanks

[matthiasblaesing]

The reason this was switched is here:

#863

The reasoning is sound from my POV - the ELF ABI is pretty clear in my interpretation: No marker == soft float.

If you know that you are on linux and hardfloat, then you can override detection by setting the system property jna.prefix to linux-arm, if you know you are on linux with softfloat calling conventions set jna.prefix to linux-armel.

Please - for questions use the google group!

[matthiasblaesing]

I'll assume this was the correct answer.