Hi, For the needs of ABRT project, I need to extract backtrace from a coredump. Currently we use gdb for this, but we want to do it without gdb.
At first I attempted to just use remote unwinding feature of libunwind for this, but quickly discovered that I can't easily implement find_proc_info(): I would need to implement DWARF parsing. Looks like the one relatively easy way forward is to extend libunwind. This way I can use its internal DWARF parsing code. This patch implements coredump unwinding using the ideas from libunwind-ptrace.h API. It adds a new directory under src, src/coredump/, and results in one new header file (libunwind-coredump.h) and one new library (libunwind-coredump.a) installed by "make install". src/coredump/ contains README file, *.c and *.h files, and example-core-unwind.c which shows how the coredump unwinding can be done. The rest of the patch tweaks automake/cofigure machinery to make src/coredump/* into build process. The patch is run-tested on i386. Due to its size of ~80kb, I send it bzip2ed to avoid exceeding mailing list limit. Sorry about it... Please review. -- vda
1.patch.bz2
Description: BZip2 compressed data
_______________________________________________ Libunwind-devel mailing list [email protected] https://lists.nongnu.org/mailman/listinfo/libunwind-devel
