[Bug symtab/28900] GDB crash on loading symbols form openbios binary

["aburgess at redhat dot com" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=28900

--- Comment #11 from Andrew Burgess <aburgess at redhat dot com> ---
With commit 336125713fc GDB should no longer crash, but this is achieved by
just ignoring the symbols stabs data that we don't currently know how to
handle.

If you do 'set complaints 1000' before loading the offending binary into GDB,
then you get to see GDB complain about all the debug symbols it can't handle. 
The output looks like:

  During symbol reading: function `_savegpr_14' appears to be defined outside
of all compilation units
  During symbol reading: function `_save32gpr_14' appears to be defined outside
of all compilation units
  During symbol reading: function `_savegpr_15' appears to be defined outside
of all compilation units
  ... snip ...
  During symbol reading: function `_restgpr_14' appears to be defined outside
of all compilation units
  During symbol reading: function `_rest32gpr_14' appears to be defined outside
of all compilation units
  During symbol reading: function `_restgpr_15' appears to be defined outside
of all compilation units
  ... snip ...

Which look like register save/restore hooks, but I don't claim to really
understand.  I don't believe that these symbols have and DWARF information
available, so these functions will not show up in something like 'info
functions', but there is a symbol for these functions in the symbol table,
which GDB does read, so you can do things like 'x/10i _savegpr_14`, however,
the symbol table entry has no type or size, so GDB doesn't understand the
bounds of the function.

Given the original bug description is for GDB crashing, and this is now fixed. 
I'd like to tentatively propose we close this bug.  If Glenn feels like there
is still important debug information that isn't being loaded, this might be
better represented in a new bug.  Given the lack of GDB developers interested
in improving stabs support (right now), I suspect any new bug is not likely to
get much attention.  Thoughts?

-- 
You are receiving this mail because:
You are on the CC list for the bug.