fix(build): libe3.pc export and relocatability

[Thecave3]

Two libe3.pc generation bugs. The first is #71; the second was found while verifying the fix for it, and has the same shape — a .pc confidently describing a location that holds no libe3.

VERSION 0.1.2 → 0.1.3, folded into the second commit.

---

1. LATREC_DEFAULT_DIR was exported as a quoted macro — closes #71

pkg-config's Cflags tokenizer strips unescaped quotes, so a consumer received the bare path and the preprocessor could not lex it:

error: expected expression before ‘/’ token

Only present when configured with LIBE3_ENABLE_LATREC=ON and LIBE3_BUILD_TESTS=ON — the latter being that option's own default — so anyone who did not think to pass -DLIBE3_BUILD_TESTS=OFF got a libe3.pc that broke their build.

No quoting fix exists. The issue offered either fixing the quoting or dropping the path; the quoting route was tested and cannot work:

form in libe3.pc	pkg-config --cflags emits	verdict
-DD="/a/b"	-DD=/a/b	quotes stripped → fails to lex
-DD='/a/b'	-DD=/a/b	same
-DD=\"/a/b\"	-DD=\"/a/b\"	survives pkg-config…

…but that last form then reaches a plain cc $(pkg-config --cflags libe3) as an unterminated literal — missing terminating " character. No single form is safe for both consumption styles, so a string-valued macro should not be exported at all.

PRIVATE loses nothing. latrec_open_in() applies the fallback from src/core/latrec.c, libe3's own translation unit, so every caller of latrec_tls_open_as() inherits the compiled-in default without seeing the macro. Nothing in tests/, examples/ or benchmarks/ references it; the one example that opens a ring (examples/sm_simple/simple_service_model.hpp:209) goes through latrec_tls_open_as(); and latrec.h:461 already has an #ifndef fallback to /tmp/latrec for anyone calling the inline path directly. The build-tree default therefore keeps working for libe3's own tests and examples, which is all it was ever for. Exporting it was wrong on a second count regardless: it is a build-tree path, meaningless to a consumer once installed.

LIBE3_ENABLE_LATREC stays PUBLIC — a bare flag with no quotes to mangle, and a consumer must see it or its own latrec_* calls resolve to the no-op stubs while linking a library that has the ring registry compiled in.

The .pc generator already failed loudly on a generator expression; it now does the same for a quoted value, so a future PUBLIC string macro cannot reintroduce this silently.

2. libe3.pc baked the configure-time prefix

libe3.pc.in wrote prefix=@CMAKE_INSTALL_PREFIX@, so the installed file described where the build expected to go rather than where it ended up. cmake --install --prefix <other> moves every file but cannot rewrite the .pc, and neither can a DESTDIR-staged packaging step. The consumer gets -I, -L and smdir under the original prefix — and pkg-config reports success while doing it, so the failure surfaces later as a missing header from a directory nobody asked for.

Reproduced on this tree: installing to a scratch prefix produced a .pc still claiming prefix=/usr/local, with pkg-config --variable=smdir libe3 naming a service-model directory that did not exist while the real one sat under the install prefix.

The prefix is now derived from ${pcfiledir}, which pkg-config expands to the directory holding the .pc. The CMake package config has always been relocatable through @PACKAGE_INIT@ — this only makes the pkg-config path agree with it.

GNUInstallDirs permits any component directory to be absolute; those have nothing relative to derive and keep their absolute value, and an absolute libdir also leaves the prefix with no relative form (that is where the .pc lands), so it falls back to the absolute prefix. The component loop also reads CMAKE_INSTALL_DATAROOTDIR rather than CMAKE_INSTALL_DATADIR, whose cache entry is empty until GNUInstallDirs derives it.

Verification

Built with LIBE3_ENABLE_LATREC=ON and LIBE3_BUILD_TESTS=ON — the configuration that used to break:

• Exported Cflags: -I${includedir} -DLIBE3_ENABLE_JSON -DLIBE3_ENABLE_ASN1 -DLIBE3_HAS_ZMQ=1 -DLIBE3_ENABLE_LATREC — no quotes left to strip.
• build/CMakeFiles/libe3.dir/flags.make still carries the private LATREC_DEFAULT_DIR, so the internal default is intact.
• Installed to a non-default prefix and consumed from there: --cflags, --libs and --variable=smdir all resolve under that prefix; a program compiles and links through the shell-split cc $(pkg-config ...) path; the reported smdir really contains sm_simple; a DESTDIR-staged install is relocatable too.
• ctest: 23/23 passed. scripts/check_commit_trailers.py: passes for both commits.

[github-actions]

CI report — dad8089 — ✅ all checks passed

Workflow	Result	Time	Run
Commit policy	✅ success	1m48s	#100
E2E dApp Integration	✅ success	2m14s	#112
E2E Topologies (multi-dApp / multi-RAN)	✅ success	2m43s	#110
Full-loop Latency Benchmark	✅ success	1m16s	#111
Unit Tests	✅ success	4m38s	#138
latrec portability	⏭️ not triggered (paths filter)	—	—
MPMC Queue Benchmark	⏭️ not triggered (paths filter)	—	—

E2E Topologies (multi-dApp / multi-RAN)

zmq/ipc

• ✅ 1 RAN - 1 dApp: indications=5
  • dapp peer=t11 ran=ran-solo sub=1 indications=5 seq=[0..4] dropped=0 (0%) age_ms(avg=0 max=0 @seq=0) hist[<=1:5 2-5:0 6-10:0 >10:0]
• ✅ 1 RAN - 2 dApps: dApp#1 ind=5 sub=1, dApp#2 ind=6 sub=2, RAN saw 2 dApps
  • dapp1 peer=t12 ran=ran-shared sub=1 indications=5 seq=[0..4] dropped=0 (0%) age_ms(avg=1 max=1 @seq=0) hist[<=1:5 2-5:0 6-10:0 >10:0]
  • dapp2 peer=t12 ran=ran-shared sub=2 indications=6 seq=[0..5] dropped=0 (0%) age_ms(avg=0.833333 max=1 @seq=0) hist[<=1:6 2-5:0 6-10:0 >10:0]
• ✅ 2 RANs - 1 dApp: from ran-a ind=5, from ran-b ind=5
  • dapp peer=t2a ran=ran-a sub=1 indications=5 seq=[0..4] dropped=0 (0%) age_ms(avg=0.2 max=1 @seq=0) hist[<=1:5 2-5:0 6-10:0 >10:0]
  • dapp peer=t2b ran=ran-b sub=1 indications=5 seq=[0..4] dropped=0 (0%) age_ms(avg=0 max=0 @seq=0) hist[<=1:5 2-5:0 6-10:0 >10:0]

zmq/tcp

• ✅ 1 RAN - 1 dApp: indications=5
  • dapp peer=default ran=ran-solo sub=1 indications=5 seq=[0..4] dropped=0 (0%) age_ms(avg=0 max=0 @seq=0) hist[<=1:5 2-5:0 6-10:0 >10:0]
• ✅ 1 RAN - 2 dApps: dApp#1 ind=5 sub=1, dApp#2 ind=6 sub=2, RAN saw 2 dApps
  • dapp1 peer=default ran=ran-shared sub=1 indications=5 seq=[0..4] dropped=0 (0%) age_ms(avg=1 max=1 @seq=0) hist[<=1:5 2-5:0 6-10:0 >10:0]
  • dapp2 peer=default ran=ran-shared sub=2 indications=6 seq=[0..5] dropped=0 (0%) age_ms(avg=1 max=1 @seq=0) hist[<=1:6 2-5:0 6-10:0 >10:0]
• ✅ 2 RANs - 1 dApp: from ran-a ind=5, from ran-b ind=5
  • dapp peer=default ran=ran-a sub=1 indications=5 seq=[0..4] dropped=0 (0%) age_ms(avg=1 max=1 @seq=0) hist[<=1:5 2-5:0 6-10:0 >10:0]
  • dapp peer=off100 ran=ran-b sub=1 indications=5 seq=[0..4] dropped=0 (0%) age_ms(avg=14.8 max=73 @seq=4) hist[<=1:4 2-5:0 6-10:0 >10:1] [note: indication queueing >1ms]
  • ⚠️ dapp indication queueing exceeded 1ms (report-only)

E2E dApp Integration

✅ posix/ipc

• dApp exit: 0
• Indications received: 7

✅ posix/tcp

• dApp exit: 0
• Indications received: 7

✅ zmq/ipc

• dApp exit: 0
• Indications received: 7

✅ zmq/tcp

• dApp exit: 0
• Indications received: 7

Full-loop Latency Benchmark

Full-loop latency

Full-loop latency benchmark (N=1102 after 50 warmup)

All values in microseconds (μs). Link: zmq, transport: ipc, encoding: ASN.1 APER.

#	Description	Tags	mean	p50	p99	max
1	Collect indication data	RECORD_BEGIN to ENCODE_E3SM_BEGIN	0.12	0.10	0.35	0.60
2	Create & encode indication	ENCODE_E3SM_BEGIN to ENCODE_E3SM_DONE	1.10	1.07	3.36	3.80
3	Encode E3AP (indication)	EMIT_ENTER to ENQUEUE, then DEQUEUE to ENCODE_E3AP_DONE	2.61	2.90	3.80	7.50
4	Queuing (indication)	ENQUEUE to DEQUEUE	5.32	5.51	15.90	83.69
5	Delivery (indication)	ENCODE_E3AP_DONE to SEND_DONE	0.28	0.25	0.73	6.29
6	E3 wire (RAN -> dApp)	SEND_DONE to RECV	53.68	54.39	67.03	452.77
7	Decode E3AP (indication)	RECV to DECODE_E3AP_DONE	1.74	1.48	2.60	23.13
8	libe3 dispatch (indication)	DECODE_E3AP_DONE to DELIVER_BEGIN	0.10	0.10	0.15	0.42
9	Decode indication	DELIVER_BEGIN to DECODE_E3SM_DONE	0.64	0.57	0.91	1.34
10	Process data	DECODE_E3SM_DONE to ENCODE_E3SM_BEGIN	0.04	0.04	0.05	0.15
11	Create & encode control	ENCODE_E3SM_BEGIN to ENCODE_E3SM_DONE	0.36	0.35	0.55	0.75
12	Encode E3AP (control)	EMIT_ENTER to ENQUEUE, then DEQUEUE to ENCODE_E3AP_DONE	3.84	3.83	5.34	6.78
13	Queuing (control)	ENQUEUE to DEQUEUE	16.76	16.34	21.65	51.72
14	Delivery (control)	ENCODE_E3AP_DONE to SEND_DONE	5.53	5.26	8.48	20.03
15	E3 wire (dApp -> RAN)	SEND_DONE to RECV	54.31	54.59	63.37	650.26
16	Decode E3AP (control)	RECV to DECODE_E3AP_DONE	2.93	2.85	6.25	15.65
17	libe3 dispatch (control)	DECODE_E3AP_DONE to DECODE_E3SM_BEGIN	0.29	0.28	0.51	0.90
18	Decode & handle control	DECODE_E3SM_BEGIN to DECODE_E3SM_DONE	0.45	0.41	0.70	0.95
Total	Total round-trip		150.68	150.26	178.75	746.76

ubuntu-latest, Release build, ZMQ + IPC, ASN.1 APER.

These numbers are measured inside a GitHub Actions container and should be treated as an upper bound on E3AP's and the library's own latency, not a representative deployment measurement.

Ready to merge (fast-forward only)

A maintainer can land the reviewed commits with:

git fetch origin
git checkout main && git merge --ff-only dad8089f2fe1ca6f895b80efff42e10fb10bfa89 && git push origin main

Head: dad8089f2fe1ca6f895b80efff42e10fb10bfa89 (branch 71-libe3pc-exports-latrec_default_dir-as-a-quoted-macro-which-pkg-config-strips-breaking-the-build). If --ff-only fails as non-fast-forward, the branch must be rebased on the latest main.

One comment per PR, rewritten in place once every workflow for dad8089 finished.