reproducible build and yocto sdk

reproducible build and yocto sdk

[Kenth Eriksson]

[-- Attachment #1: Type: text/plain, Size: 1520 bytes --]

I'm having issues with reproducible builds on kirkstone when comparing a build driven by bitbake compared to a build driven from a yocto sdk. Can I expect reproducible build between bitbake builds and sdk builds? Some produced binaries have minor differences, but larger c++ binaries show quite huge diff. One common difference is that the ELF header seems to have different OS/ABI between the two builds. Can binutils on different host give such a result?

diffoscope

│┄ File has been modified after NT_GNU_BUILD_ID has been applied.
├── readelf --wide --file-header {}
│ @@ -1,20 +1,20 @@
│  ELF Header:
│ -  Magic:   7f 45 4c 46 01 01 01 03 00 00 00 00 00 00 00 00
│ +  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 - GNU
│ +  OS/ABI:                            UNIX - System V
│    ABI Version:                       0
│    Type:                              DYN (Position-Independent Executable file)
│    Machine:                           ARM
│    Version:                           0x1
│    Entry point address:               0x1f04
│    Start of program headers:          52 (bytes into file)
│ -  Start of section headers:          41316 (bytes into file)
│ +  Start of section headers:          41328 (bytes into file)


Thanks,
Kenth

[-- Attachment #2: Type: text/html, Size: 4510 bytes --]

Re: [bitbake-devel] reproducible build and yocto sdk

[Richard Purdie]

On Mon, 2023-05-29 at 09:53 +0000, Kenth Eriksson via
lists.openembedded.org wrote:
> 
> I'm having issues with reproducible builds on kirkstone when
> comparing a build driven by bitbake compared to a build driven from a
> yocto sdk. Can I expect reproducible build between bitbake builds and
> sdk builds? Some produced binaries have minor differences, but larger
> c++ binaries show quite huge diff. One common difference is that the
> ELF header seems to have different OS/ABI between the two builds. Can
> binutils on different host give such a result?

This is probably an OE-Core question rather than a bitbake one.

Offhand, it looks like the linker is using --hash-style=gnu for one and
--hash-style=sysv for the other.

Cheers,

Richard