HSF Packaging Group Meeting #14, November 15, 2017
Present: Ben Morgan, Chris Green, Dario Menasce, Graeme Stewart, Patrick Gartung, Martin Ritter, Emil Obreshkov, Javier Cervantes Villanueva, Raphal Pacholek, Guilherme Amadio
- Next meeting on the 29th November
- Graeme has booked room 40-R-D10 at CERN.
- Call for volunteers to give talks on their projects/experiment needs.
- Specifically on Use Cases and on the contents of the Test Stack.
- Offer to get external speaker on Portage. Maybe not next meeting but beyond.
- Looking at speakers from ATLAS and CVFMS.
- Went through the Google Doc
- Ben M had some discussion with Andrea Valassi that we may want to follow up on (HEP_OSLibs meta RPM to define
- Proposed that we maybe start with a reduced package std - notes taken directly in the document.
- Dario - we do need to foresee scaling up the complexity.
- We anticipate that individual experiments will build their own stacks - our main goal is to provide a tool to do that with ease and flexibility.
- ABI incompatibilities we will need to be capable of dealing with, but don’t need to worry too much about it right now. (Chris - clang is definitely having issues in the last few versions).
- Patrick - we have the HSF spack repo so most of these should be covered (some were upstreamed). Guilherme thinks most of this is also described in portage.
- build system into a container would be a first step.
- Ben M - demonstrate from scratch source/binary install
Use Cases for Packaging
- Went through the Google Doc
- Graeme motivated Use Cases
- HSF Packaging Tech Note enumerated required
features but not so clear on the use cases that lead to these.
- Not much about on installation and runtime use cases.
- Some offline discussion with WG/users on the above to get started.
- Draft Google Doc split
into sections to cover
- Build: how the package is built from source into binary
- Install/Deployment: how/where the binary products are distributed for end use
- Development: how a user sets up their environment to use the products
- Item 3.c (incremental builds) in Build section considered very important in LHCb and ATLAS.
Example use case(s) of reusing both system packages and packages from another “layer”,
in this case the LCG stack.
- Patrick - similar in CMS, but more use of rpms to define reuse
- Suggested that implementing a GUI to track which recipes were used would be useful
- Comment that containers could help to define a runtime environment.
- Additional discussion mentioned issue of “one container per release” leading
to duplication. Possibility to mount local or other container filesystem in container
might mitigate this.
- Question on how to handle Debug symbols in a package stack. Comment that traditionally
there’s a separate stack of Debug packages, but does duplicate things. Should
look at possibility of stripping Debug symbols and storing somewhere, as Linux
- Call for input and comments on Google Doc