FAQ Feedback
Please add questions you would like to see answered in the following FAQs!
FAQ QUESTIONS
What are reproducible builds?
- -> Link to definition?
- see Defining Reproducible Builds definition II session
Why should I care?
- -> Give one example for each persona (user, dev/packager, non-technical ppl) and link to buy-in for details.
- -> Integrate info from Use cases session
Is there any reproducible distribution that I can use today?
- Sadly we are not there yet, but multiple distributions are putting a lot of effort to reach that goal.
What is the current status of reproducible builds?
- [answer missing]
I am interested in making my software reproducible - FAQ
When did the reproducible builds project start?
Some of the earliest public work on reproducible builds started in Bitcoin (2011) and TorBrowser (2013) .
How do I make my software reproducible?
Check that your software is reproducible (see [#check]), then look at the differences and modify your build to avoid this difference. A number of general guidelines are available in the “Get deterministic builds” section of the documentation.
Is it possible?
Yes it is possible and actually rather simple by following some best practices. see (tips & helps in https://reproducible-builds.org/docs/)
How do I check that my software is reproducible?
Build your software once. Add a variation to the build environment which should not influence the output, for example by moving it to another path and create a second build. Then compare the resulting binaries using the @diffoscope@ tool. Make sure there are no differences. See the test bench documentation for a number of parameters that can be varied.
Are there common mistakes to avoid?
In general
During the build, many different programs might be executed, for example compilers, code generators and other scripts. To ensure that the output is reproducible, each program should have a deterministic behavior and not leak information about the system into the produced output artifacts (e.g. program binaries).
For a C program
- Avoid
__FILE___
. - Avoid
__DATE__
,__TIME__
, and__TIMESTAMP__
(you can also disable them using the flag -D in gcc/clang). - Do not read uninitialized memory (msan, asan) /!\ not in openssl /!\ (if your tool is part of a toolchain)
- Ensure to not rely on readdir() order (if your tool is part of a toolchain) - e.g. similar to how
find | cpio -o needs
a sort in the middle
For a Python program
-
Python does hash randomization. Anything that depends on the value of a hash is not deterministic (e.g. iterating over a hash table). Setting the environment variable PYTHONHASHSEED can disable this behavior, but the proper fix is not to depend on specific hash values.
-
Pyc files
- you may also simply discard .pyc and .pyo files
For emacs byte code
- elc files
How do I make compression tools produce reproducible output?
gzip
- use “gzip -n”
ziptime
resets timestamps in ZIP files to 2008-01-01, meant for Android APKs
ar
- Use option “D” (“Operate in deterministic mode”).
multi-threaded producers (compression, compilers etc)
Most multi-threaded software provides non predictable output which results in non reproducibility. Forcing them to run mono/single-threaded ensures the output is predictable.
I’m interested in verifying the reproducibility of software I use - FAQ
(this section might not have any answerable questions right now…)
I’m interested in packaging/distributing software in a reproducible way. - FAQ
I’m a developer of a distribution and would like to make my distribution reproducible. - FAQ
- sign up for the the mailing list:
- talk to us on irc: irc.oftc.net #reproducible-builds
- link to documentation:
- link to tools:
- https://reproducible-builds.org/tools/ ** update to be non-debian-specific links
- documented known issues, bugs, etc. (currently very debian, plans to generalize for cross-distro)
I’m interested in creating a build farm that tests the reproducibility of my distribution/project - FAQ
- how to best run the tests (parallelized?)
- example of some imperfect helper scripts https://github.com/bmwiedemann/reproducibleopensuse
- where to report results, in what format?
- how frequent should tests be?
- Which parts of the build environment should be varied between builds? See the test bench documentation
Common concerns about reproducible builds
When I debug I’d like to see the full path of the sources my binary was built from
- (Why? maybe for editing source files before compiling and testing again)
- SOURCE_MAP_… PATH/PREFIX? is optional?
I really need to know where and when the binary used by users has been built
- this could be tracked in metadata outside of the shipped binary - e.g. a database mapping hash-of-binary to actual build time and place or a (build-log-)file acompanying the binary
- once your software builds fully reproducibly, meaning you get the same binary built on every host at any time, the values do not matter anymore
How to get involved?
- how to help various projects?
- how to help with documentation?
- how to help with outreach
- how can I sponsor the effort?
OTHER QUESTIONS
Instructions: Please add answers to the below questions
What are the benefits of reproducible builds?
- see also Use cases session or rather the refined outcome from it
What are yet unsolved problems in reproducible builds?
UNSORTED FAQ QUESTIONS
Instructions: If you have a question you’d like to see in a Reproducible FAQ, but don’t know where to add it, please write the question here.
set up website mirror on github:
- create account on github for automatic pushes from alioth
- setup post-receive hook to push to github mirror
- include username/password of auto-push account
- push to https://github.com/reproducible-builds/website
- https://www.chiliproject.org/projects/chiliproject/wiki/howto_mirror_your_git_repository_on_github