[Lldb-commits] [PATCH] D108061: [lldb] Add support for shared library load when executable called through ld.

[Greg Clayton via Phabricator via lldb-commits]

clayborg added a comment.

In D108061#2955707 <https://reviews.llvm.org/D108061#2955707>, @rdhindsa wrote:

> Looking further as to why executable symbols are not loaded, it looks like when we read the link_map in lldb, AddSOEntries() is called for all the entries in link_map read from ld.so. The first entry's name(which corresponds to executable) is empty, which seems to be set by ld.so here (https://github.com/bminor/glibc/blob/master/elf/rtld.c#L1389). main_map at this location has the entry for respective executable, but its name is removed, hence lldb while going through entries can't read entry about main executable. 
> I verified that lldb does iterate through the link_map list with this patch, reading that file name as empty. I also verified that if respective line is commented in ld.so , and hence the name of executable is retained, lldb with this patch is able to set breakpoint on main as well. Hence, it would require change to glibc as well. Once that is in, we can add additional test for that feature later on as a follow-up patch. Does that sound okay?

Since the executable name is empty, I wonder what would happen if we looked up the address in the memory region information. This information is populated by the /proc/<pid>/maps file, and each memory region has a name, and that name is usually the path to the object file on disk. As one last effort can you check if the address' memory region info has the path to the executable?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D108061/new/

https://reviews.llvm.org/D108061