Enhancement-605: a .probe card in a deck that opens with a .control block

Enhancement-605: a .probe card in a deck that opens with a .control block

Scope: src/frontend/inpc_probe.c (probe_head: the head insertions placed past
a leading .control block; deck no longer moved onto the inserted card),
examples/probeblock_examples/ (new, 5 checks per solver). ngspice only. Finding
N3 of the 2026-09-10 integration hunt
(docs/bug_hunts/2026-09-10_ngspice-osdi-integration-second-hunt.md),
a stock ngspice defect that every OSDI deck with a .probe met.

Suites: probeblock_examples 5 of 5 per
solver, both solvers; probeshort and the other .probe suites unchanged; full sweep
502 of 502 (one sweep, Enhancements 604–608 folded together).

What was wrong

.control
pre_osdi rc.osdi
.endc
...
.probe v(out)

.save: no such command available in ngspice

inp_probe adds a .save all card when the deck has no save of its own, and
inserted it right after the first card after the title. A deck that opens with a
.control block — every OSDI deck does, for pre_osdi — got the card inside that
block, where it ran as a command and was lost; the probe still worked through the
vectors the probes themselves register, so the line read as a spurious error. The
.save line the probes generate for their vectors, and a differential probe's source,
went to the same place. And deck was moved onto the inserted card, so the walks that
follow started one card late.

What changed

• probe_head(deck): a line meant for the head of the deck goes after the .endc of
  a .control block that opens it, and after the first card otherwise — at all three
  head-insertion sites.
• deck stays on the first card.

Verification

check	result
a leading .control block, then .probe v(out)	no ".save: no such command"; 0.5 V printed
...with .probe i(r1) v(out)	r1#branch 0.5 mA
an OSDI deck (pre_osdi block first) with .probe alli	quiet, complete
no leading block, the probed device on the deck's first line	unchanged
the deck's own .save beside a .probe, leading block	the generated .save line past the block too

Full sweep 502 of 502 on both solvers.

Write-up: enhancements_doc/Enhancement-605.md