bneradt commented on PR #13892: URL: https://github.com/apache/trafficserver/pull/13892#issuecomment-6450865944
Thanks for the detailed re-review. Updated in 64d94bec62 (folded into the existing commits). - The tracked launcher is no longer an AC_CONFIG_FILES output in either build mode. Only urtest-configured.sh is generated; out-of-tree builds get a portable launcher copy. Automake's CONFIG_CLEAN_FILES now contains only the configured wrapper. New regressions run configure, tests/distdir, and tests/distclean in both layouts, checking that the distributed launcher matches the source and that cleanup preserves the source. I also verified the actual full source tree's in-tree configure and tests/distdir/distclean. This is not a claim that the unrelated top-level make dist failure is fixed. - Included only the async-engine wait_cleanup use-after-free hunk from #13013. - Resolved the replay audit root before relative-path comparisons and added a symlinked-checkout regression. - Corrected the null-transform description: it checks the expected response and diagnostics, not proof that ATS synthesizes Content-Length. Validation: Autotools ASAN build/install and make -j8 format succeeded; all 124 framework tests passed. The TLS engine, TLS engine abort, and null-transform tests passed under ASAN, with two additional successful abort runs (3/3 total). I agree the inherited abort client does not deterministically exercise teardown while an async job is pending. I corrected its misleading documentation, but have not closed that coverage gap or weakened its existing assertions. A passing abort run must not be treated as proof that removing async_ep.stop() is caught. The previously verified HTTP/2 synchronization is unchanged. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
