| Summary: | JUnit sig-segvs when attempting to print a stack backtrace | ||
|---|---|---|---|
| Product: | frysk | Reporter: | Andrew Cagney <cagney> |
| Component: | general | Assignee: | Andrew Cagney <cagney> |
| Status: | RESOLVED FIXED | ||
| Severity: | normal | ||
| Priority: | P1 | ||
| Version: | unspecified | ||
| Target Milestone: | --- | ||
| Host: | Target: | ||
| Build: | Last reconfirmed: | ||
| Project(s) to access: | ssh public key: | ||
| Bug Depends on: | |||
| Bug Blocks: | 1580 | ||
|
Description
Andrew Cagney
2006-02-10 04:58:49 UTC
The problem here is that the java runtime, and frysk are fighting over the SIGCHLD handler. Even though frysk doesn't create any child processes, GCJ does. GCJ can create two processes: - a process to demangle symbols - [possibly] a process to perform address -> line conversion and install a null sigchld handler. Frysk, in turn, installs its own long-jumping handler. A crash occures when frysk's handler finds itself processing a signal that was ment for the core. It attempts to longjmp through a bad buffer. This helps, the jumpbuf has a flag indicating that it should be used. It will stop the bad long-jump. 2006-02-09 Andrew Cagney <cagney@redhat.com> * cni/Poll.cxx (struct poll_jmpbuf): Replace poll_env. A per-thread buffer. Index: frysk-core/common/ChangeLog 2006-02-10 Andrew Cagney <cagney@redhat.com> * Makefile.gen.sh (echo_LDFLAGS): Append $(GEN_GCJ_NO_SIGCHLD_FLAGS). (echo_name_): New function. (echo_LDFLAGS): Use. (echo_MANS): Check if man page is needed. (echo_PROGRAMS): Move call to echo_MANS out of echo_PROGRAMS. (GEN_UBASENAME): Rename GEN_BASEUNAME). * Makefile.rules (GEN_GCJ_NO_SIGCHLD_FLAGS): Define. Index: frysk-core/frysk/event/ChangeLog 2006-02-10 Andrew Cagney <cagney@redhat.com> * TestSigChild.java: New. unfortunatly this makes stack backtraces sux |