[rb-general] Request for help to make mariadb-10.3 in Debian reproducible

[Chris Lamb]

Hi Otto,

> Unfortunately none of the changes I made seemed to solve this..
> 10.3.22 is still unreproducible in unstable due to RocksDB, TokuDB and
> Mroonga.

Mmm, alas, I only managed to get this to "almost reproducible" stage
before my attention was required elsewhere.

However, I quickly note that we are still embedding the build flags
that vary depending on the build path via -ffile-prefix-map into the
binary, despite my patch that was designed precisely to normalise them:

│ │ │ │ │    0x00012040 2d676e75 27202743 464c4147 533d2d67 -gnu' 'CFLAGS=-g
│ │ │ │ │    0x00012050 202d4f32 202d6666 696c652d 70726566  -O2 -ffile-pref
│ │ │ │ │ -  0x00012060 69782d6d 61703d2f 6275696c 642f3173 ix-map=/build/1s
│ │ │ │ │ -  0x00012070 742f6772 6f6f6e67 612d392e 312e323d t/groonga-9.1.2=
│ │ │ │ │ +  0x00012060 69782d6d 61703d2f 6275696c 642f322f ix-map=/build/2/
│ │ │ │ │ +  0x00012070 67726f6f 6e67612d 392e312e 322f326e groonga-9.1.2/2n

I would suggest that this curiosity is the first place to start. :)

§

> The upside is that all upstreams seemed pretty responsive. If there
> just was some very easy instructions for upstreams to follow on how to
> build their project twice with the variations reproducible-builds.org
> does, they could be able to study the actual binaries themselves more
> easily.

I understand and feel your pain. That tool ought to "just" be reprotest
but as previously discussed I don't possess the ability to become the
maintainer for this.


Best wishes,

-- 
      ,''`.
     : :'  :     Chris Lamb
     `. `'`      lamby at debian.org 🍥 chris-lamb.co.uk
       `-