Add version information to /sample/info endpoint

[corneliusroemer]

I was hoping I could get the version of LAPIS that's serving a response from the /sample/info endpoint but it turns out I can't.

Might be nice to add this:

This came up as part of testing new versions of LAPIS and SILO on staging.pathoplexus.org and wondering whether I was already getting responses from the new versions (without having to log in to argocd, which of course works but takes more clicks)

[fengelniederhammer]

I also thought about adding this, but unfortunately it isn't that simple:

• for a released version SILO and LAPIS could simply read its version from a file and return it
• for builds on branches, we should not display a version number, but probably
  • the commit hash when running a Docker image that was built on CI
  • "local" when running it locally.

[Taepper]

Actually, we always have the version.txt file in the project root (automatically updated by release please).

Then the branches would return the "latest released version"

[Taepper]

We might use the approach from GenSpectrum/LAPIS#1019 for SILO running in Docker images

[pflanze]

I suggest to use the following script:

#!/bin/bash
set -meuo pipefail
IFS=

header_path=include/exact_version.h

# Take version from current commit if in the git checkout, otherwise
# fall back to the version from release please.
v=$(git describe --tags --match 'v*' 2> /dev/null || { echo -n v; cat version.txt; })

printf '#define PROGRAM_VERSION "%q"\n' "$v" > "$header_path"

# Prevent the new mtime of the file from leading to spurious rebuilds:
touch -d 1972/01/01 "$header_path"

Add the path include/exact_version.h to .gitignore, and run the above script as first step during the build (I don't know cmake so am not useful here).

[pflanze]

And include the file in the right place of course. PS. add a #pragma once too:

printf '#pragma once\n\n#define PROGRAM_VERSION "%q"\n' "$v" > "$header_path"

[Taepper]

Thank you very much!

git describe --tags --match 'v*' 2> /dev/null seems to really do exactly what we want. The rebuilding problem due to mtime is already solved by just using a preprocessor definition. I also believe it is less complex as we do not need to dynamically generate a file. I pushed a draft which shows how to do it with cmake to #679