C call graph: calls through a #define alias (failf -> Curl_failf) are not linked

Open
  • Is something wrong?
#7,828 1 comment 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

Nobody has claimed this yet.

Assessment

Difficulty
4/5
Estimated time
3-5 days
Newbie friendliness
56/100
Issue type
Bug
Clarity
Mostly clear
Activity status
Active
Tech stack
c, python
Domain
devtools

Research direction

Start with the C call graph resolution path and the callgraph_triage.py reproducer described in the issue. Inspect how calls and object-like macro definitions are represented, then compare results against the libclang reference on the curl 7.30.0 corpus. Done means calls through function-alias macros link to their function targets, including the described guarded aliases, and call graph recall improves without introducing incorrect confident links.

Written by the indexing model from the issue text.

Description

bug

What's wrong. In C, a call written through an object-like macro alias isn't linked to the function it names. curl 7.30.0 has #define failf Curl_failf at lib/sendf.h:48 (infof likewise). Calls written failf(data, ...) aren't linked to Curl_failf: lib/ftp.c has 43 such sites, imap.c 19, url.c 18, smtp.c 16, and so on. libclang resolves them through the macro. They make up most of the 293 not_extracted/other edges callgraph_triage reports for that snapshot, where call recall vs libclang 18.1.1 is 65.4%. Precision is unaffected: 0 wrong confident links.

Repro. Run tests/tools/callgraph_triage.py c with corpus_repos pointed at a curl-7_30_0 checkout, then read the not_extracted/other samples, e.g. lib/connect.c:1103 failf(data, "Connection time-out");.

Suggested fix shape. In C, an object-like macro whose whole body is one identifier naming a function defined in the repository is an alias: resolve a call to the macro as a call to that function. A guarded alias (curl's infof also has Curl_nop_stmt / (void) variants under #if) links from the branch that names a function. If the engine doesn't evaluate #if, it links to every branch's function target.

Impact. Under-counted call edges, so function fan-in, function PageRank and cross-file fcall edges are low for logging/error helpers in older C code bases that alias them. Modern curl calls Curl_failf directly, and its recall is 86.4%.

How it was found. The same temporal-crucible validation as #7827 (the include issue): gitgalaxy's own call_graph_resolution._score and callgraph_triage, run with the libclang reference on curl and nDPI snapshots. Write-up: squid-protocol/temporal-crucible branch validate/engine-c, docs/engine_validation.md.

🤖 Generated with Claude Code

Dominant language
Python
Stars
57
Forks
11
Avg merge
1h 35m
Merged PRs (30d)
975

Getting set up

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from squid-protocol/gitgalaxy

All issues in squid-protocol/gitgalaxy

Similar issues

More Python issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.