| Summary: | It should be possible to disable kernel module package splitting | ||
|---|---|---|---|
| Product: | [Build System, Metadata & Runtime] OE-Core | Reporter: | Ola Nilsson <ola.x.nilsson> |
| Component: | configuration | Assignee: | Bruce Ashfield <bruce.ashfield> |
| Status: | RESOLVED WONTFIX | QA Contact: | |
| Severity: | enhancement | ||
| Priority: | Low | CC: | bruce.ashfield, randy.macleod |
| Version: | unspecified | ||
| Target Milestone: | Future | ||
| Hardware: | All | ||
| OS: | Multiple | ||
| Whiteboard: | |||
| OS type for building Yocto: | --- | Type of Regression: | --- |
| Verified: | Documentation change: | Yes (doc changes required) | |
|
Description
Ola Nilsson
2017-10-10 09:33:53 UTC
Whilst this kind of change may improve rootfs generation time, the common path for most people will still be doing module splitting to limit the number of packages actually installed. By making this kind of change, if it is even possible will increase the testing requirement for a non-common use case. We implemented this internally @ Wind River in the past, and it ended up not being put into production. It did speed up packaging and rootfs generation, but it adds complexity to an already complex process and loses some features that are generally valuable (i.e. the ability to upgrade a single module, rootfs size increase, etc). Any implementation of this still has to RPROVIDE all of the existing module package names, since there are many recipes that request individual modules that now need to pull in the unified kernel module package. After consultation, the complexity of the multiple kernel image builds, the module split logic, rdepends and the speedups with rpm4 and parallel installation removing the module splitting at this time (even optionally) is not something we are planning to do. I've resolved this as 'won't fix', but if there's a strong feeling to park it as a future/enhancement, I can do that as well. |