Repository navigation
Depend on the monolithic idemix v0.2.0 module - #510
Conversation
|
fabric-ca is an application, distributed as a binary. It is not intended to be consumed as a Go library. Rationalising the imports is fine but it must remain exactly compatible with Fabric. |
| github.com/IBM/idemix v0.0.2-0.20240913182345-72941a5f41cd | ||
| github.com/IBM/idemix/bccsp/types v0.0.0-20240913182345-72941a5f41cd | ||
| github.com/IBM/mathlib v0.0.3-0.20250709075152-a138079496c3 | ||
| github.com/IBM/idemix v0.1.1 |
There was a problem hiding this comment.
There looks to be a v0.2.0 release now.
There was a problem hiding this comment.
Good catch, will move to idemix v0.2.0.
| github.com/IBM/idemix/bccsp/types v0.0.0-20240913182345-72941a5f41cd | ||
| github.com/IBM/mathlib v0.0.3-0.20250709075152-a138079496c3 | ||
| github.com/IBM/idemix v0.1.1 | ||
| github.com/IBM/mathlib v0.2.0 |
There was a problem hiding this comment.
There is now a v0.3.0 release, which is a dependency of github.com/IBM/idemix@v0.2.0
There was a problem hiding this comment.
Will bump mathlib to v0.3.0 to match idemix v0.2.0.
Thanks. For context, we don't run the fabric-ca binary — we embed the client library (lib.Client + api) for programmatic per-institution enroll/register, which the CLI can't cover. Fully agree it must stay in lockstep with Fabric. The trigger here: fabric-token-sdk (LFDT Panurus) and fabric-smart-client moved to the monolithic github.com/IBM/idemix, which can't coexist with fabric-ca's split bccsp/types in one build (ambiguous import). Single-module is also the current idemix direction — v0.1.0/v0.1.1 and today's v0.2.0 are all single-module, while Fabric is still on the 2024 split commit. So fabric-ca will need this realignment whenever Fabric refreshes idemix; this just does it early. |
Require github.com/IBM/idemix v0.2.0 (monolithic) instead of the split bccsp/types and bccsp/schemes/weak-bb sub-modules, and bump mathlib to v0.3.0 (a dependency of idemix v0.2.0). Both the split sub-modules and the monolithic module export the github.com/IBM/idemix/bccsp/types import path, so fabric-ca cannot build in any module graph that also pulls the monolithic idemix (ambiguous import). The consolidated module folds the sub-packages back into a single module; fabric-ca builds unchanged against it. Signed-off-by: Evan <evanyan@sign.global>
8aaf76e to
3ed5da6
Compare
|
Tick the box to add this pull request to the merge queue (same as
|
Type of change
Description
fabric-ca currently depends on the split idemix layout —
github.com/IBM/idemix v0.0.2-...plus the standalonegithub.com/IBM/idemix/bccsp/typesandgithub.com/IBM/idemix/bccsp/schemes/weak-bbmodules.Both the split sub-modules and the monolithic
github.com/IBM/idemixmodule export thegithub.com/IBM/idemix/bccsp/typesimport path. As a result, any module graph that pulls fabric-ca alongside a dependency that requires the monolithic idemix fails to build with an ambiguous import, e.g.:This blocks using the fabric-ca client library in the same build as recent
fabric-token-sdk/ LFDT Panurus releases, which depend on the monolithicgithub.com/IBM/idemix v0.1.1.This change depends on the monolithic
github.com/IBM/idemix v0.1.1and drops the splitbccsp/typesandbccsp/schemes/weak-bbrequirements;github.com/IBM/mathlibis bumped tov0.2.0to match. v0.1.1 folds the sub-packages back into the single module, so thegithub.com/IBM/idemix/bccsp/...import paths still resolve.Additional details
No source changes — only
go.mod/go.sumand the regeneratedvendor/tree. fabric-ca compiles unchanged against idemix v0.1.1; verified locally withgo build ./....Related issues
N/A