On 2024/05/15 13:12:44 Christopher Schultz wrote: > All, > > A recent thread was posted with a tcnative crash with not much in the > way of useful information in the error: > > https://lists.apache.org/thread/m1dbj3w1x1oqftqsbj7jbnvkm2073x1o > > The error details were: > > " > # EXCEPTION_ACCESS_VIOLATION (0xc0000005) at pc=0x00000001800ccd10, > pid=1244, tid=0x0000000000000ab0 > # > # JRE version: OpenJDK Runtime Environment (Zulu 8.60.0.21-CA-win64) > (8.0_322-b06) (build 1.8.0_322-b06) > # Java VM: OpenJDK 64-Bit Server VM (25.322-b06 mixed mode windows-amd64 > compressed oops) > # Problematic frame: > # C [tcnative-1.dll+0xccd10] > # > # Core dump written. Default location: D:\Program > Files\apache-tomcat\bin\hs_err_pid1244.mdmp > " > > So, not super helpful unless you happen to have a debugger handy. > > If we had a debug build available for users, we should be able to get > better information coming back from that failure, possibly a complete > native back-trace. > > IIRC, building a debug version just involves adding something obvious > like /DEBUG to the compiler and/or linker and/or NOT stripping-out the > debug symbols after the build is complete. > > Would this represent a burden on the release manager to produce both > kinds of builds for an official release?
I would be in favor of this. For that I have tried it yesterday. Produced a debug build of OpenSSL and libtcnative. MSVC has added to new DLLs to the runtime path and I have no idea where to get them, it does not load. One of it is msvcrtd.dll. Are we actuall able to supply that to users otherwise it does not make sense to provide a debug build. M --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
