iOS release: fix erl_errno shim newline + preflight the driver table

[GenericJam]

Two issues that break mix mob.release --ios against the Elixir 1.20-rc.5 / OTP 29 runtime (OTP hash 7d46fdd4). Both only affect the App Store release path — the dev build is unaffected — which is why they hadn't surfaced before (AirCartMax / Code to Cloud shipped on an older OTP). Found while shipping a Livebook-on-mob app.

1. erl_errno_id_unknown shim written with a literal \n

The weak-stub line was printf '%s\\n' '…' > erl_errno_id_compat.c inside the ~S""" (raw, no-escape) release-script heredoc. The two backslashes reach bash verbatim, so printf '%s\\n' emits a literal backslash-n into the C file and clang rejects the trailing }\n:

erl_errno_id_compat.c:1:101: error: expected identifier or '('
    1 | …return "unknown"; }\n

Fix: printf '%s\n' (one backslash) so a real newline is emitted. This matches the working SLIM-log printf elsewhere in the same heredoc. Added a regression guard to release_script_test (assert sh =~ ~S|printf '%s\n'…|, refute sh =~ ~S|printf '%s\\n'|).

2. Cryptic failure when priv/generated/driver_tab_ios.c is missing

The iOS release links a per-app static-NIF driver table from priv/generated/driver_tab_ios.c, but the dev build uses Mob's built-in Zig table and mix mob.regen_driver_tab defaults to Zig. A project that only ever did dev builds (or ran regen without --format c) dies deep in release_device.sh with cc: no such file or directory: 'priv/generated/driver_tab_ios.c'.

Fix: a build_ipa preflight (check_driver_table/0) that fails early via the existing {:error, _} path with the exact command to run (mix mob.regen_driver_tab --format c). Chosen over auto-regen because --format c tables are meant to be hand-editable, so silently regenerating could clobber edits.

Verification

• mix test test/mob_dev/release_script_test.exs → 35 passed (incl. new guard).
• Both fixes verified end-to-end producing a signed, App-Store-accepted IPA.

Not included

A third issue — the OTP 29 iOS libbeam.a (hash 7d46fdd4) shipping both erl_posix_str.o and erl_errno_str.o, each defining _erl_errno_id (ld64 duplicate-symbol; zig/LLD tolerates it) — lives in the OTP cross-build / published tarball, not here. Tracked separately.

🤖 Generated with Claude Code