(long_options): Add "debug" and "linker".
(main): Handle -d and -l arguments. Make command line input and
output files optional. Parse the command file before opening the
BFD's, which requires storing more information in local variables.
If INPUT names multiple files, link them together. Use OUTPUT for
the output file name if not named on command line.
(show_usage): Changed for new options.
(link_inputs): New function to automatically invoke linker to
handle multiple INPUT files.
(choose_temp_base_try, choose_temp_base, pexecute): New functions,
mostly copied from gcc/gcc.c.
* nlmconv.h (input_files, output_file): Declare.
* nlmheader.y (input_files, output_file): Define.
(command): Support INPUT with a string_list argument. Support
OUTPUT.
(string_list): Renamed from module_list.
* Makefile.in (nlmconv.o): Define LD_NAME based on
program_transform_name.
Fixes PR 3974.
(copy_file): Don't copy EXEC_P specially here.
This permits strip to generate an executable file on UnixWare;
it still does not actually strip the debugging information.
alignment adjustment. Clear BSF_SECTION_SYM if symbol is moved to
a different section. Use time_t for time variable.
(setup_sections): Only put sections with contents in output NLM.
(i386_mangle_relocs): No symbols are common at this point. Add
casts to avoid warnings.
(main): Set the bss vma to always follow the data section. Move
symbols into new sections, and adjust values by output_offset.
(setup_sections): Don't copy all sections, but instead point all
text sections to .text, all data sections to .data, and all bss
sections to .bss.
(copy_sections): Adjust accordingly.
* configure.in: If we have some *-*-netware* target, or are using
all targets, set BUILD_NLMCONV to $(NLMCONV_PROG) in Makefile.
* Makefile.in (PROGS): Use $(BUILD_NLMCONV) rather than
$(NLMCONV_PROG).
(dump_headers): Don't test for SEC_BALIGN if it's not defined by bfd.h.
(The latter is done with #ifdef in case it turns out SEC_BALIGN has to be
put back. I didn't *see* any uses of it besides this one, though.)
sections before looking at the symbols. Move all common symbols
into .bss section. Automatically define _edata and _end. Only
export symbols in the export list, with multiple prefixes if
necessary. Warn if no version. Always create extended header.
Set date automatically if not already set.
(setup_sections): Count size of bss sections.
(mangle_relocs, i386_mangle_relocs): Accept section argument, and
take reloc_count as a changeable pointer; changed callers.
(i386_mangle_relocs): Remove PC relative relocs within a section,
as they require no adjustment.
* nlmheader.y: Fixed memory allocation throughout: token STRING is
now allocated on the heap, and freed if not needed. Null
terminated copyright message. Accept version with only two
strings.
(symbol_list_opt): New nonterminal, either symbol_list or empty.
(symbol_list): Use left recursion to avoid overflowing parser
stack.
(yylex): Rearranged beginning of line check. Accept quoted
strings using single quotes. End generic argument at comment
character or parentheses.
(string_list_append): Fixed.
(string_list_append1): New function.
(bsssize, datasize, textsize): New global variables.
(bss_section_name, data_section_name, text_section_name): Removed.
(print_berkeley_format): Map berkeley_sum over all the sections,
rather than only reporting sizes of specifically named sections.
* Makefile.in ($(OBJDUMP_PROG)): Removed dependency on size.o.
* objdump.c (remove_useless_symbols): New function.
* objdump.c (comp): Simplify.
* objdump.c (dis-assemble_data): Make simpler and more
efficient how we filter out useless symbols: Just filter
BEFORE the sort (using remove_useless_symbols).
* objdump.c (objdump_print_address): Simplify.
Change output syntax to match gdb.
