* ldgram.y: Add token LNAME.
(input_list): Treat LNAME like NAME, but pass it to
lang_add_input_file as lang_input_file_is_l_enum.
* ld.texinfo: Document using -lFILENAME in INPUT.
bfd_check_format_matches, and, if ambiguous, print a list of
matching formats. If file format is not recognized, treat file as
a linker script.
* ldgram.y (yyerror): If assuming an object file is a script,
mention that. Tweak the format of the error messages.
* ldlex.l (lex_warn_invalid): If assuming an object is a script,
guess that this is not actually a script, and just report that the
file format was not recognized.
* ld.texinfo (Options): Admit that -( may be used more than once.
Add note that unrecognized object files are now treated as linker
scripts.
* ldfile.c (ldfile_input_filename): Make const.
(ldfile_assumed_script): New variable.
(try_open): Change arguments types to const.
(ldfile_find_command_file): Likewise.
(ldfile_open_command_file): Likewise. Also, set lineno to 1.
* ldfile.h: Update declarations for ldfile.c changes.
* ldlex.l: Include <ctype.h>.
(file_name_stack): Change to be const char *.
(lineno_stack): New static variable.
(<<EOF>>): Set lineno as well as ldfile_input_filename.
(lex_push_file): Make name argument const. Initialize
lineno_stack entry.
(lex_redirect): Initialize lineno_stack entry.
(lex_warn_invalid): Handle non printable characters nicely.
* ldlex.h (lex_push_file): Declare second argument as const.
* ldgram.y (ifile_p1): Recognize GROUP.
* ldlex.l: Recognize GROUP.
* ld.texinfo (Option Commands): Document GROUP.
* lexsup.c (S_ISDIR): Define if not already defined.
(parse_args): Add support for -rpath. If -R is used to name a
directory, treat it as -rpath for Solaris compatibility.
* emultempl/elf32.em (gld${EMULATION_NAME}_before_allocation):
Pass command_line.rpath to bfd_elf32_size_dynamic_sections.
* ldmain.c (main): Initialize command_line.rpath to NULL.
* ld.texinfo: Document -rpath option.
* lexsup.c (parse_args): Add traditional_format to longopts, and
handle it.
* ldmain.c (main): Initialize config.traditional_format to false.
* ldlang.c (ldlang_open_output): Set BFD_TRADITIONAL_FORMAT in BFD
flags of output_bfd according to config.traditional_format.
* ld.texinfo: Document -traditional-format.
* ldlang.c (out_bfd_get_section_by_name): Remove.
(wild_section): Call bfd_get_section_by_name rather than
our_bfd_get_section_by_name. Don't call wild_doit if there is no
section.
(lang_create_output_section_statements): Remove.
(map_input_to_output_sections): For several cases, call init_os if
it has not already been called.
(lang_size_sections): If output section was not created, skip it.
(lang_process): Don't call lan_create_output_section_statements.
(lang_place_orphans): Skip files with just_syms_flags set to true.
* ld.texinfo: Document change.
* ld.texinfo (Options): Document changes to option syntax.
* Makefile.in: Update dependencies.
* ldver.c (help): Tweak dashes in usage message.
* ldgram.y (%union): Remove unused members.
Remove %tokens for command line options; add ones for input types.
(command_line): Rules removed.
(file): Instead of command line, recognize an
input type indicator, then use the nonterminal for that type.
(defsym_expr): New nonterminal from code formerly in command_line.
* ldlex.h: Declare parser input type enum and variable.
Don't declare parse_line.
* ldlex.l: Remove unused variables. Make some used ones static
and comment them.
(COMMAND): Start state and its rules removed.
At start of yylex, return input state token if at start of input.
(lex_redirect): Don't need to set yyout.
(ldlex_command): Function removed.
* ldmain.c (main): Instead of calling parse_line, set up the
redirections and call yyparse directly.
* ldmisc.c (vfinfo): If there's no input filename, print nothing, not
"command line".
* lexsup.c: Remove #if 0'd code.
(parse_line): Function removed.
(parse_args): Rewrite to use getopt_long_only.
(set_default_dirlist): New function from code formerly in
ldgram.y:command_line.
(set_section_start): New function.
* emultempl/generic.em, emultempl/gld960.em, emultempl/hppaelf.em,
emultempl/lnk960.em, emultempl/m88kbcs.em: Don't enclose
compiled-in link scripts in "{" and "}", as the grammar no longer
wants them to be.
(1) show `-y SYMBOL' with space
(2) weaken multi-target stuff, to avoid false expectations now that
linker now often config'd with only one target
(3) describe -relax support on i960
(4) correct description of hyphen chars in symbols
(5) describe CONSTRUCTORS linker script cmd
It wasn't documented (or likely used) and wastes time.
(try_open): If EXTEN is empty, don't try it.
* ldctor.c, lderror.c, ldexp.c, ldfile.c, ldindr.c, ldlang.c,
ldlex.l, ldmain.c, ldmisc.c, ldsym.c, ldver.c, ldwarn.c,
ldwrite.c, lexsup.c, mri.c, relax.c: Replace DEFUN macro calls
with normal function declarations.
* Move *.em to emultempl/*.em. Move *.sh to emulparams/*.sh.
Move *.sc-sh to scripttempl/*.sc.
* {emultempl,emulparams,scripttempl}/README: New files.
* sh.em, st2000.em, z8ksim.em, h8300hms.em, h8500hms.em: Files
removed, replaced with generic.em.
* h8300.sh, h8500.sh, h8300.sc, h8500.sc: Renamed from
h8[35]00hms.s[ch]. Change their contents to omit the "hms".
* *.em (*_get_script): Return script name instead of script contents.
* ldlang.c (lang_process): Change caller.
* ldlex.l, ldgram.y: Recognize -m option.
Check for input files after *all* options in grammar.
* ldmain.c (main): Check for -m options. Add default directory
for -m.
* mkscript.c: File removed.
* genscripts.sh: Take two more parameters, tooldirlib and libdir,
to add to the default LIB_PATH.
Look for input files in the new subdirectories.
Create the scripts in emulations subdirectory and don't filter
them through mkscript.
* configure.in: Make the emulations subdirectory.
* Makefile.in: Account for all of the above changes.
Remove unused .SUFFIXES. Get libgcc.a path with gcc
-print-libgcc-file-name instead of $(libdir)/libgcc.a.
Put CFLAGS last in the compilation rules.
Add -I../bfd to INCLUDES so sysdep.h is found.
* ldfile.c (try_open): If opening without the extension fails,
try with the extension even if -v or -V was given.
had_script is imported (from ldgram.y), not exported.
(2) restructure machine dependencies into separate chapter
(3) replace pretty, but quick to become obsolete, graph of BFD platforms
vs architectures with extended excerpt from (and ref to ) objdump -i.
description believed complete as of today; still holes---small
ones---in "commands" section. BFD section also needs better table of
supported archs and formats; see FIXMEs for scattered q's and
desiderata.
