Note that T5996 has been backported and a full explanation is to be found in the previous comment. Close as resolved, with the needed workaround for earlier images.

This was a worthwhile investigation, with the goal of using Cython as a method of easily wrapping the standard netlink headers. Although workable, it appeared more trouble than it's worth, as one still has to cheat to navigate e.g. pass by reference, and limitations of Cython at the time. A much better approach would be a standard Python extension module ... set to 'on hold' for reference and until we have a plan for development, as has been recently discussed.

As c-po mentioned above, this was an interesting idea that was not workable; it was subsumed by the investigations in T3876.

@tjh in fact, this was a design decision, discussed with @Viacheslav at the time, and agreed upon, although neither of use are fully satisfied with the decision (and I'll let him amend these claims, if needed):

Checking a recent 1.5 build (1.5-rolling-202405090225), I do not see the discrepancy mentioned in the description: the result for 1.5 is also (just) missing vif entries, and one can observe that the syntax error in vif 2 causes cascading syntax errors in the other vif sections.
The following script recreates the effect of the XorpConfigParser:

Migration requires the config file to be in the standard syntax, such as produced by the system itself (set/del; commit; save) --- in this case, a configuration that can be loaded and saved on VyOS 1.3.5, would not be missing braces as above. If one needs to hand edit a config file, at the very least one needs to load/save on the source system before migration. The only weird corner case, linked above, is when a change of mac address triggers a rewrite by the legacy parser; that case is also restored by saving the config before migration.

This will be possible after subtask T6006, for 1.5; backport can be considered at that time, but for now limit tags to Circinus.

PR:
https://github.com/vyos/vyos-1x/pull/3443

See details and explanation in subtask T6327; notable, one can configure as system console, but setting at boot has limited use.

Note that the work in the link above, but for the get_commit_schedule function itself, has already been added in subtasks T6319/T6146. As mentioned above, the get_commit_schedule function is not useful (and potentially misleading) until we replace the legacy commit algorithm in 1.5. Remove 1.4.0 tag.

PR for 1.5:
https://github.com/vyos/vyos-1x/pull/3372
Combined PRs for backport to 1.4 of T5839, T5660, T6276 pending.

Reopen to investigate.

PR:
https://github.com/vyos/vyos-1x/pull/3346

The fix mentioned above has been backported to 1.4: T5996.