Adding prism to Google OSS-Fuzz continuous fuzzing

[tranquac]

Hi Kevin and the prism team,

I've submitted a PR to add prism to Google OSS-Fuzz,
Google's continuous fuzzing infrastructure for open-source projects.

OSS-Fuzz PR: google/oss-fuzz#15339

What this adds

Two fuzz targets that run 24/7 on Google's ClusterFuzz infrastructure:

Target	What it fuzzes
fuzz_parse	Prism.parse, Prism.dump, Prism.parse_stream, Prism.parse_comments
fuzz_lex	Prism.lex, Prism.parse_lex

Seed corpus: 986 fixture files from prism's own test/prism/fixtures/ tree
(including version-specific subdirs and the seattlerb/whitequark/unparser edge cases).

Dictionary: Ruby keywords and operators from prism's own fuzz/dict.

Why this matters

prism is Ruby's official parser — fuzzing it continuously with AddressSanitizer and
UndefinedBehaviorSanitizer can surface memory safety issues (buffer overflows,
use-after-free, integer overflows) before they reach production.

What the OSS-Fuzz reviewers need

The reviewers typically ask the primary project maintainer to leave an LGTM comment on
the PR. If you're open to this integration, a comment on
google/oss-fuzz#15339 would be very helpful.

Happy to answer any questions or adjust the integration if you have suggestions.

Thanks!

[Earlopain]

Prism is also a C library. Why go through the ruby extension instead of just using that? Presumably you can enable the memory sanitizer then, and remove all the other exceptions that ruby needs

[Earlopain]

cc @stevenjohnstone since this is probably interesting to you

[tranquac]

Thanks @Earlopain — that's a great point. I've revised the integration to fuzz libprism directly as a standalone C library instead of going through the Ruby extension.

Updated PR: google/oss-fuzz#15339

Changes:

• language: c (was language: ruby)
• Base image: base-builder (was base-builder-ruby) — Ruby installed only for templates/template.rb generation
• Build: make static CC=$CC CFLAGS=$CFLAGS → link against build/libprism.a via $LIB_FUZZING_ENGINE
• Sanitizers now include memory sanitizer (msan) — not possible with the Ruby extension approach
• Removed all Ruby VM ASAN_OPTIONS workarounds
• Two targets: fuzz_parse (uses upstream fuzz/parse.c) and fuzz_lex (new fuzz_lex.c targeting pm_serialize_lex())

cc @kddnewton — your LGTM on the OSS-Fuzz PR would help move this forward for review.