> Here is an updated version of the patch > <http://lists.debian.org/debian-boot/2008/02/msg00208.html>, with > additional fixes and split into pieces as requested. I do not believe > there are any dependencies between these pieces. > > Part 1: Uninitialized data fix, as in previous patch:
applied
> Part 2: Endian fix, as in previous patch:
applied
> Part 3: Correct diagnostic text when symbol cannot be found, new:
>
> Index: src/mklibs.py
> ===================================================================
> --- src/mklibs.py (revision 51573)
> +++ src/mklibs.py (working copy)
> @@ -469,7 +493,7 @@
> for name in needed_symbols:
> if not name in symbol_provider:
> if not needed_symbols[name].weak:
> - raise "No library provides non-weak %s" % symbol
> + raise "No library provides non-weak %s" % name
My python is not particularly good, but I don't see where symbol is
defined. Only way it could be set is in one of the loops above, but
that would have no relation to the name.
> Part 4: Support running mklibs for a set of binaries including gdbserver
> (which defines symbols left unresolved in libthread_db), new:
>
> Index: src/mklibs.py
> ===================================================================
> --- src/mklibs.py (revision 51573)
> +++ src/mklibs.py (working copy)
> @@ -394,8 +414,12 @@
> libraries = set(force_libs)
> for obj in objects.values():
> for symbol in undefined_symbols(obj):
> - debug(DEBUG_SPAM, "needed_symbols adding %s, weak: %s" %
> (symbol, symbol.weak))
> - needed_symbols[str(symbol)] = symbol
> + # Some undefined symbols in libthread_db are defined in
> + # gdbserver, not in any library.
> + if (not (re.search("libthread_db\.so", obj)
> + and re.search("^ps_", str(symbol)))):
> + debug(DEBUG_SPAM, "needed_symbols adding %s, weak: %s" %
> (symbol, symbol.weak))
> + needed_symbols[str(symbol)] = symbol
> libraries.update(library_depends(obj))
>
> # calculate what symbols are present in small_libs and available_libs
The code looks ok, but I don't know how we feel about adding
library/program specific hacks into mklibs.
> Part 5: Sysroot and multilib support, as in the previous patch but with an
> additional fix (-L options pointing to sysroot directories that would be
> searched anyway are bad since they can cause the linker to treat the
> directories as non-sysroot directories):
applied
--
see shy jo
signature.asc
Description: Digital signature
