You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Segmentation fault when running codebase-memory-mcp cli index_repository on dhw/xx_hash.h #355
Thanks for the report. I tried to reproduce this and couldn't with the headers available to me: I added a regression guard that runs the vendored xxhash.h (~7,500 lines, same macro-dense family as xx_hash.h) through C extraction under AddressSanitizer, and it extracts cleanly on current main.
Since your report, several C/C++ extraction crash fixes have landed (NULL-terminated param_types OOB #322, an expression-eval step budget #323, and template-substitution NULL handling #360), so this class of segfault is likely already resolved.
Could you help confirm?
Does the crash still reproduce on the latest build (v0.7.0 / current main)?
If yes, could you attach (or link) the exact dhw/xx_hash.h that crashes? With the file I can add a precise regression test and fix the specific path.
Thanks for the report — the segfault was in C extraction of a heavily-macro'd header (dhw/xx_hash.h), which is the same crash family fixed by the C-LSP hardening that landed for v0.7.0:
Your exact dhw/xx_hash.h isn't public, so I added a proxy regression test (clsp_nocrash_issue355_xxhash_header, commit ba49137) that drives the vendored xxhash header (~7.5k lines, same macro-dense family) through cbm_extract_file and asserts no crash. It passes on v0.7.0.
Given the fixes target exactly this crash class and the representative macro-dense header now extracts cleanly, I'm closing this as resolved in v0.7.0. If your specific dhw/xx_hash.h still segfaults on v0.7.0, please reopen and attach the file (or a minimal reduction) — I'll add it as a direct regression test and dig in. Thanks!
Activity
DeusData commentedon May 30, 2026
Thanks for the report. I tried to reproduce this and couldn't with the headers available to me: I added a regression guard that runs the vendored
xxhash.h(~7,500 lines, same macro-dense family asxx_hash.h) through C extraction under AddressSanitizer, and it extracts cleanly on currentmain.Since your report, several C/C++ extraction crash fixes have landed (NULL-terminated
param_typesOOB #322, an expression-eval step budget #323, and template-substitution NULL handling #360), so this class of segfault is likely already resolved.Could you help confirm?
main)?dhw/xx_hash.hthat crashes? With the file I can add a precise regression test and fix the specific path.I'll keep this open pending your reply. 🙏
test(c-lsp): xxhash-header extraction guard (issue #355 proxy)
DeusData commentedon May 31, 2026
Thanks for the report — the segfault was in C extraction of a heavily-macro'd header (
dhw/xx_hash.h), which is the same crash family fixed by the C-LSP hardening that landed for v0.7.0:Your exact
dhw/xx_hash.hisn't public, so I added a proxy regression test (clsp_nocrash_issue355_xxhash_header, commit ba49137) that drives the vendored xxhash header (~7.5k lines, same macro-dense family) throughcbm_extract_fileand asserts no crash. It passes on v0.7.0.Given the fixes target exactly this crash class and the representative macro-dense header now extracts cleanly, I'm closing this as resolved in v0.7.0. If your specific
dhw/xx_hash.hstill segfaults on v0.7.0, please reopen and attach the file (or a minimal reduction) — I'll add it as a direct regression test and dig in. Thanks!