Tools such a JLINK_Server fail to find SVD files in local installed packs

[ReinhardKeil]

Describe the bug

Tools such a JLINK_Server have in some cases problems to find SVD files as the current directory depends on actions that the user may have performed.

To Reproduce
The problem happen in some caese when a the M55_HP core is launched.

Proposed solution

For local pack repositories use absolute paths in the *.cbuild-run.yml file. In this case:

    - file: c:/w/alif_ensemble-cmsis-dfp/Debug/SVD/AE822FA0E5597BS0_CM55_HE_View.svd

[brondani]

cbuild-run may reference several local files, such as output (elf/hex) and dbgconf. Such files are relative to cbuild-run itself regardless of current/working directory.
For this reason it sounds inconsistent to emit absolute paths only for svd files.
I believe the svd path should be resolved where it is relevant for the jlink_server command line construction, maybe around getSvdFilePaths in vscode-cmsis-debugger:
https://github.com/Open-CMSIS-Pack/vscode-cmsis-debugger/blob/6dd709ae30d89aeedb63fc1b6158836d122956c0/src/cbuild-run/cbuild-run-reader.ts#L46
@jreineckearm Can you please comment on this matter?

[jreineckearm]

I can see the concern, but the big difference is that output and dbgconf files are workspace specific and expected to be close by. Whereas local packs are less likely to be.

SVD files normally have entries of the form:

    - file: ${CMSIS_PACK_ROOT}/vendor/pack/1.0.0/devices/device.svd
      type: svd
      pname: cm33_core0

Hence, for local packs printing the full svd path could be equally considered consistent.
And it has the advantage of some degree of portability of workspaces if a development team agrees on a "local pack installation location".

vscode-cmsis-debugger may not be involved in headless/CI runs using pyocd only. Hence, resolution there would only help VS Code users in interactive sessions. But purpose of cbuild-run is to configure such runs, too.

[brondani]

@TeoMahnic
For pyOCD from the code snippets below it seems you are already correctly resolving relative references in cbuild-run, including svd file references, can you please confirm?

https://github.com/pyocd/pyOCD/blob/162dac4fce43c8d1ef6e6c9d761a5cc14440f80f/pyocd/target/pack/cbuild_run.py#L228-L230
https://github.com/pyocd/pyOCD/blob/162dac4fce43c8d1ef6e6c9d761a5cc14440f80f/pyocd/target/pack/cbuild_run.py#L914

[TeoMahnic]

Yes, pyOCD resolves relative paths in cbuild-run based on the location of the cbuild-run file itself, rather than the current working directory.