Mailinglist Archive: opensuse-factory (463 mails)

< Previous Next >
[opensuse-factory] openSUSE reproducible builds status 2019-06

last month's status:

Last months' reproducible builds project updates (including my work):

This month, due to recent rebuilds, the number of verified packages grew
so much that I could check that currently all verifiable Tumbleweed
binaries contain what they should.
Not verified are:
1: 348+ packages known to not build reproducibly
2: 16 efi and kmp packages - mostly because was reverted
3: ~80 packages using _multibuild without empty FLAVOR
(needs work in my tooling)
4: ~5 packages with unsubmitted local fixes - that number shrunk a lot
last month

The other caveat here is that this is using build-compare and that
already had 3 bugs that caused it to report different rpms as identical.
To avoid the possibility of such bugs in the future,
I'd like to see bit-identical verifiable rpms for Leap 16. This is
tracked in

Newly reproducible core packages:

had unreproducible pdf from inkscape - file was dropped

only reproducible when building with %do_profiling 0
That is not the default in Factory to gain 8% compiler performance.
note: gcc9 is similar.

maybe from rcc toolchain fixes? or maybe there are still low-entropy
order issues in rcc output


Still unreproducible:

I uploaded today
and rbstats (with comparison to last month) are:
total-packages: 12014 (+103)
build-tried: 12004 (+95)
build-failed: 48 (-5)
build-n-a: 106 (-5)
build-succeeded: 11850 (+105)
build-official-failed+na: 173 (-9)

build-compare-failed: 348 (+10)
build-compare-succeeded: 11502 (+95)

verify-failed: 531 (-1184)
verified-semi-reproducible: 11246 (+1309)
verified-bit-identical: 0 (+0)

bit-by-bit-identical: 11189 (+84)
not-bit-by-bit-identical: 651 (+15)
shows the change over time
lists very unreproducible core packages (bootstrap+DVD)

Of the badly unreproducible packages,
3 were in ring0
39 were in ring1

That makes it 42/2901 => 1.45 %
which is below the overall average of
348/11850 => 2.94 %

651/11850 => 5.49 % of packages are not perfectly reproducible

Bernhard M.

< Previous Next >
This Thread
  • No further messages