docs: restore the pn_ parameter names in registrar, mid_registrar and…

[denys-i-didww]

Summary

Commit e941906 (docs: fix README.md typos, 5 Aug 2026) replaced every occurrence of
pn with on in the registrar and mid_registrar documentation. That commit fixed a great
many real typos across 135 files; in these pages it also rewrote identifiers, and pn
here is not a typo — it is Push Notification.

Eight module parameters in each of the two modules now carry a name that exists nowhere in
the source tree:

Documented today	Registered in lib/reg/pn.h
on_enable	pn_enable
on_providers	pn_providers
on_ct_match_params	pn_ct_match_params
on_pnsreg_interval	pn_pnsreg_interval
on_trigger_interval	pn_trigger_interval
on_skip_on_interval	pn_skip_pn_interval
on_refresh_timeout	pn_refresh_timeout
on_enable_purr	pn_enable_purr

The parameters come from one macro in lib/reg/pn.h, which is why the same eight are
affected in both modules. A configuration written from the current pages does not start:
OpenSIPS refuses an unknown modparam while parsing, before any module is initialised.

What else the substitution reached

• the async function pn_process_purr(), documented as on_process_purr();
• the RFC 8599 Contact URI parameters, in prose and as the default value of
  pn_ct_match_params: pn-provider, pn-prid, pn-param became on-provider, on-prid, on-param, and ;pn-purr=XXX became ;on-purr=XXX;
• the abbreviation itself — "Push Notification (PN) support" now reads "Push
  Notification (ON) support", and "PN-enabled contact" reads "ON-enabled contact";
• one cross-reference in usrloc, which cites registrar's pn_trigger_interval.

on_skip_on_interval is the clearest fingerprint: both halves of pn_skip_pn_interval
were replaced, including the one in the middle of the name.

Reproduced

On the official opensips/opensips:4.0 image, one modparam per run, opensips -C:

$ opensips -C -f t.cfg            # a name the README carries today
ERROR:core:set_mod_param_regex: parameter <on_enable> not found in module <registrar>
ERROR:core:parse_opensips_cfg: bad config file (1 errors)

$ opensips -C -f t.cfg            # the name this PR restores
(exit 0)

All sixteen documented names are refused, eight in each module, with the same message.
All sixteen restored names are accepted.

pn_enable needs one note: set to 1 it declares a dependency on tm and
event_routing (lib/reg/pn.c:227-238), so a config without event_routing stops on
that instead — with the module loaded it starts, and so does pn_enable=0. The swept
name on_enable is still refused as not found with event_routing loaded, so the name
is the variable and the dependency is not.

The async function was checked the same way, in a script that calls it: pn_process_purr()
loads, on_process_purr() fails to parse.

Scope

This PR restores pn only. Everything else that commit changed in these files is left as
it is — overwride → overrides and the rest are real corrections and are not touched.

Verification

After this change every parameter documented in the two modules matches a name registered
in lib/reg/pn.h — the eight above, checked by name against the macro. Nothing else was
rewritten: t_on_reply() and t_on_branch() still read as before in the registrar
examples, and no on_, on- or ON token that belongs there was touched.

Not in this PR

The same commit renamed two more identifiers elsewhere, and they need their own decision
rather than being folded in here:

• httpd: the documented callback httpd_acces_handler_cb became
  httpd_access_handler_cb. The typo is real, but it is a C API symbol declared in
  httpd_load.h:68 and used at httpd.c:240, so the fix belongs in the code, not in the
  page describing it.
• db_virtual: db_max_consec_retrys became db_max_consec_retries. Same shape — the
  code carries the misspelling, and the correct fix is in the code. Happy to send either.

---

AI assistance disclosure. This report and the change were produced with AI assistance.
Every item was checked against the source before submission:

• the eight names are registered in lib/reg/pn.h:90-97, in a macro shared by both
  modules;
• no on_* form of any of them appears in any .c or .h file in the tree —
  repository-wide grep;
• pn_process_purr is the async function checked for at lib/reg/pn.c:211 and :217;
• pn-provider, pn-prid, pn-param are the literals at lib/reg/pn.c:54-56, and the
  default value of the parameter at lib/reg/pn.c:44;
• the failures and the successes were run, not inferred — sixteen parameters in two
  modules, plus the async function, on the official 4.0 image; the commands and their
  output are quoted above;
• the substitution was traced to a single commit — git log -S on each name over the full
  history, unshallowed for this purpose — rather than assumed from the current state;
• the two identifiers listed under Not in this PR were found by the same pass over all
  135 files that commit touched.