Automatically downloads and creates path_names lists for different Socrates versions using new Socrates github - #301
Draft
sit23 wants to merge 3 commits into
Draft
Conversation
Socrates was closed-source until now; the Met Office have open-sourced it at github.com/MetOffice/socrates under BSD-3-Clause. Registering it as a submodule lets GitHub's dependency graph show the relationship and gives users a standard way to fetch the matching source (`git submodule update --init`), instead of the old manual download from the Met Office Science Repository Service. The submodule is intentionally left uninitialized by a plain clone (git's default submodule behaviour) since the upstream repo is ~290MB, almost all of it example/data files Isca's build never touches. The codebase.py auto-fetch path added in a later commit performs a sparse partial clone of just src/ and make/ instead, which is what most users will actually hit.
… setup Socrates is now pinned as a git submodule and fetched automatically (sparse partial clone) on first compile if not already present, replacing the old workflow of manually downloading a release and pointing GFDL_SOC at a single hardcoded `trunk` checkout -- the same design whose dangling symlink crashed a trip_test run when GFDL_SOC pointed at a missing directory. GFDL_SOC_DIR now lets multiple versions be cached and takes priority over the legacy single-version GFDL_SOC override. The hand-maintained per-version path_names file, which broke on every Socrates release as files moved or were renamed, is replaced by isca.socrates_paths: it derives the needed file list from Socrates' own Makefile manifests plus a dependency-closure scan seeded from Isca's interface code. A handful of tested versions are committed at src/extra/model/socrates/socrates_version_paths/<version>; any other version is generated on the fly. Validated by real compiles (not just the generator), which surfaced and fixed three bugs along the way: - an INTERFACE block's restated call signature was miscounted as a second definition of the callee, making it look ambiguous and silently defeating the check that excludes files depending on out-of-scope code - Socrates' own aux/seaalbedo.f is both an independent manifest entry and textually INCLUDEd wholesale by aux/seaalbedo_driver.f, so compiling both duplicated every symbol seaalbedo.f defines - auto-fetching the default (submodule-pinned) version placed a plain symlink at the path .gitmodules registers as a submodule, which corrupts `git status`/`git diff` for the whole superproject; that version now always goes through a real (filtered, sparse) `git submodule update --init` instead Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
GitHub Actions' default checkout doesn't run `git submodule update --init`, but still leaves an empty directory at the submodule's path (unlike a bare `git clone` run manually, which was what local testing used) -- ensure_socrates_source() treated that as an unexpected/invalid state and raised, instead of recognising it as safe to populate. Separately, aux/seaalbedo_driver.f inherits a real module dependency (spline_evaluate_mod) from aux/seaalbedo.f's content via a plain Fortran INCLUDE. mkmf's automatic dependency detection only scans each compiled file's own literal USE/MODULE statements, never the text of files it INCLUDEs, so that dependency is invisible to it and seaalbedo_driver.o's build order relative to spline_evaluate.o is unconstrained -- whether a single `make` pass succeeds depends on incidental scheduling (confirmed by reproducing both a pass and a fail from the identical checkout). Worked around with an opt-in best-effort `make -k` pass before the normal one, scoped to SocratesCodeBase/SocColumnCodeBase via a new make_retry_keep_going flag, rather than changing compile.sh's behaviour for every other codebase. Both fixes verified against a real fresh clone (no --recurse-submodules, matching actions/checkout) through a full compile of both SocratesCodeBase and SocColumnCodeBase. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Contributor
Author
|
CI is now passing, and this now includes (for the first time) Socrates-specific compile tests, with the Socrates code downloaded from Github as part of the CI. Going to do some manual testing and then will convert to non-draft P/R. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
One of the significant problems with Isca's use of Socrates is that it has required the user to manually download Socrates themselves, and also to create a list of path_names. This P/R changes both - it uses the Met Office's new github repo for Socrates to automatically checkout Socrates if it is not already present, and then works out which path_names should be used for that version. This was built off the old socrates_versions branch of my fork, with new functionality built and tested by Claude. Claude has done a significant amount of testing of this new functionalityy, but this P/R will remain a draft until I have manually tested the functionality.