maint check-symtabs failures -- now what?

[Andrew Cagney]

> I'm chasing an elusive problem in GDB 6.1 (slightly modified, but I
> don't think that's the issue here).  The symptom is that one person's
> kernels (NetBSD plus a pile of stuff -- big file) generate these
> errors:
> 
> (gdb) p main
> $1 = {<text variable, no debug info>} 0x804372d0warning: (Internal error: pc 0x804372d1 in read in psymtab, but not in symtab.)
> 
>  <main>

- look over the hopefully still fixed symtab/1519 for clues

- check that the program only contains one type of debug info (.stabs)

Andrew


> Some functions suffer from this, some don't.  The compiler is GCC
> 3.0.1 for MIPS, with stabs.
> 
> What's odd is that most of us don't see this, but one person sees it a
> lot.
> 
> I tried "maint check-symtabs".  I figured that might help, but the
> trouble is that it spits out vast quantites of error messages.  Many
> of them seem to have to do with type names.  (The same command,
> applied to a program built with GCC 3.3.3, dwarf-2, also produces
> piles of errors -- but those seem to be ONLY for c++ mangled names.)
> 
> I've looked at the symtab stuff in gdb but found it rather opaque,
> with three different "symbol tables" that seem to contain 3 different
> sets of stuff.
> 
> So...
> 
> a. Where do I look for this?
> b. Would using dwarf-2 be expected to help?
> c. Would a newer GDB be expected to help?
> 
>    paul
> 
>