/* Generic symbol file reading for the GNU debugger, GDB. Copyright 1990, 1991, 1992, 1993, 1994, 1995 Free Software Foundation, Inc. Contributed by Cygnus Support, using pieces from other GDB modules. This file is part of GDB. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */ #include "defs.h" #include "symtab.h" #include "gdbtypes.h" #include "gdbcore.h" #include "frame.h" #include "target.h" #include "value.h" #include "symfile.h" #include "objfiles.h" #include "gdbcmd.h" #include "breakpoint.h" #include "language.h" #include "complaints.h" #include "demangle.h" #include "inferior.h" /* for write_pc */ #include "obstack.h" #include #include #include #include "gdb_string.h" #include "gdb_stat.h" #include #ifdef HAVE_UNISTD_H #include #endif #ifndef O_BINARY #define O_BINARY 0 #endif /* Global variables owned by this file */ int readnow_symbol_files; /* Read full symbols immediately */ struct complaint oldsyms_complaint = { "Replacing old symbols for `%s'", 0, 0 }; struct complaint empty_symtab_complaint = { "Empty symbol table found for `%s'", 0, 0 }; /* External variables and functions referenced. */ extern int info_verbose; /* Functions this file defines */ static void set_initial_language PARAMS ((void)); static void load_command PARAMS ((char *, int)); static void add_symbol_file_command PARAMS ((char *, int)); static void add_shared_symbol_files_command PARAMS ((char *, int)); static void cashier_psymtab PARAMS ((struct partial_symtab *)); static int compare_psymbols PARAMS ((const void *, const void *)); static int compare_symbols PARAMS ((const void *, const void *)); static bfd * symfile_bfd_open PARAMS ((char *)); static void find_sym_fns PARAMS ((struct objfile *)); /* List of all available sym_fns. On gdb startup, each object file reader calls add_symtab_fns() to register information on each format it is prepared to read. */ static struct sym_fns *symtab_fns = NULL; /* Flag for whether user will be reloading symbols multiple times. Defaults to ON for VxWorks, otherwise OFF. */ #ifdef SYMBOL_RELOADING_DEFAULT int symbol_reloading = SYMBOL_RELOADING_DEFAULT; #else int symbol_reloading = 0; #endif /* If true, then shared library symbols will be added automatically when the inferior is created. This is almost always what users will want to have happen; but for very large programs, the startup time will be excessive, and so if this is a problem, the user can clear this flag and then add the shared library symbols as needed. Note that there is a potential for confusion, since if the shared library symbols are not loaded, commands like "info fun" will *not* report all the functions that are actually present. */ int auto_solib_add_at_startup = 1; /* Since this function is called from within qsort, in an ANSI environment it must conform to the prototype for qsort, which specifies that the comparison function takes two "void *" pointers. */ static int compare_symbols (s1p, s2p) const PTR s1p; const PTR s2p; { register struct symbol **s1, **s2; s1 = (struct symbol **) s1p; s2 = (struct symbol **) s2p; return (STRCMP (SYMBOL_NAME (*s1), SYMBOL_NAME (*s2))); } /* LOCAL FUNCTION compare_psymbols -- compare two partial symbols by name DESCRIPTION Given pointer to two partial symbol table entries, compare them by name and return -N, 0, or +N (ala strcmp). Typically used by sorting routines like qsort(). NOTES Does direct compare of first two characters before punting and passing to strcmp for longer compares. Note that the original version had a bug whereby two null strings or two identically named one character strings would return the comparison of memory following the null byte. */ static int compare_psymbols (s1p, s2p) const PTR s1p; const PTR s2p; { register char *st1 = SYMBOL_NAME ((struct partial_symbol *) s1p); register char *st2 = SYMBOL_NAME ((struct partial_symbol *) s2p); if ((st1[0] - st2[0]) || !st1[0]) { return (st1[0] - st2[0]); } else if ((st1[1] - st2[1]) || !st1[1]) { return (st1[1] - st2[1]); } else { return (STRCMP (st1 + 2, st2 + 2)); } } void sort_pst_symbols (pst) struct partial_symtab *pst; { /* Sort the global list; don't sort the static list */ qsort (pst -> objfile -> global_psymbols.list + pst -> globals_offset, pst -> n_global_syms, sizeof (struct partial_symbol), compare_psymbols); } /* Call sort_block_syms to sort alphabetically the symbols of one block. */ void sort_block_syms (b) register struct block *b; { qsort (&BLOCK_SYM (b, 0), BLOCK_NSYMS (b), sizeof (struct symbol *), compare_symbols); } /* Call sort_symtab_syms to sort alphabetically the symbols of each block of one symtab. */ void sort_symtab_syms (s) register struct symtab *s; { register struct blockvector *bv; int nbl; int i; register struct block *b; if (s == 0) return; bv = BLOCKVECTOR (s); nbl = BLOCKVECTOR_NBLOCKS (bv); for (i = 0; i < nbl; i++) { b = BLOCKVECTOR_BLOCK (bv, i); if (BLOCK_SHOULD_SORT (b)) sort_block_syms (b); } } /* Make a copy of the string at PTR with SIZE characters in the symbol obstack (and add a null character at the end in the copy). Returns the address of the copy. */ char * obsavestring (ptr, size, obstackp) char *ptr; int size; struct obstack *obstackp; { register char *p = (char *) obstack_alloc (obstackp, size + 1); /* Open-coded memcpy--saves function call time. These strings are usually short. */ { register char *p1 = ptr; register char *p2 = p; char *end = ptr + size; while (p1 != end) *p2++ = *p1++; } p[size] = 0; return p; } /* Concatenate strings S1, S2 and S3; return the new string. Space is found in the symbol_obstack. */ char * obconcat (obstackp, s1, s2, s3) struct obstack *obstackp; const char *s1, *s2, *s3; { register int len = strlen (s1) + strlen (s2) + strlen (s3) + 1; register char *val = (char *) obstack_alloc (obstackp, len); strcpy (val, s1); strcat (val, s2); strcat (val, s3); return val; } /* Get the symbol table that corresponds to a partial_symtab. This is fast after the first time you do it. In fact, there is an even faster macro PSYMTAB_TO_SYMTAB that does the fast case inline. */ struct symtab * psymtab_to_symtab (pst) register struct partial_symtab *pst; { /* If it's been looked up before, return it. */ if (pst->symtab) return pst->symtab; /* If it has not yet been read in, read it. */ if (!pst->readin) { (*pst->read_symtab) (pst); } return pst->symtab; } /* Initialize entry point information for this objfile. */ void init_entry_point_info (objfile) struct objfile *objfile; { /* Save startup file's range of PC addresses to help blockframe.c decide where the bottom of the stack is. */ if (bfd_get_file_flags (objfile -> obfd) & EXEC_P) { /* Executable file -- record its entry point so we'll recognize the startup file because it contains the entry point. */ objfile -> ei.entry_point = bfd_get_start_address (objfile -> obfd); } else { /* Examination of non-executable.o files. Short-circuit this stuff. */ objfile -> ei.entry_point = INVALID_ENTRY_POINT; } objfile -> ei.entry_file_lowpc = INVALID_ENTRY_LOWPC; objfile -> ei.entry_file_highpc = INVALID_ENTRY_HIGHPC; objfile -> ei.entry_func_lowpc = INVALID_ENTRY_LOWPC; objfile -> ei.entry_func_highpc = INVALID_ENTRY_HIGHPC; objfile -> ei.main_func_lowpc = INVALID_ENTRY_LOWPC; objfile -> ei.main_func_highpc = INVALID_ENTRY_HIGHPC; } /* Get current entry point address. */ CORE_ADDR entry_point_address() { return symfile_objfile ? symfile_objfile->ei.entry_point : 0; } /* Remember the lowest-addressed loadable section we've seen. This function is called via bfd_map_over_sections. In case of equal vmas, the section with the largest size becomes the lowest-addressed loadable section. If the vmas and sizes are equal, the last section is considered the lowest-addressed loadable section. */ static void find_lowest_section (abfd, sect, obj) bfd *abfd; asection *sect; PTR obj; { asection **lowest = (asection **)obj; if (0 == (bfd_get_section_flags (abfd, sect) & SEC_LOAD)) return; if (!*lowest) *lowest = sect; /* First loadable section */ else if (bfd_section_vma (abfd, *lowest) > bfd_section_vma (abfd, sect)) *lowest = sect; /* A lower loadable section */ else if (bfd_section_vma (abfd, *lowest) == bfd_section_vma (abfd, sect) && (bfd_section_size (abfd, (*lowest)) <= bfd_section_size (abfd, sect))) *lowest = sect; } /* Process a symbol file, as either the main file or as a dynamically loaded file. NAME is the file name (which will be tilde-expanded and made absolute herein) (but we don't free or modify NAME itself). FROM_TTY says how verbose to be. MAINLINE specifies whether this is the main symbol file, or whether it's an extra symbol file such as dynamically loaded code. If !mainline, ADDR is the address where the text segment was loaded. If VERBO, the caller has printed a verbose message about the symbol reading (and complaints can be more terse about it). */ void syms_from_objfile (objfile, addr, mainline, verbo) struct objfile *objfile; CORE_ADDR addr; int mainline; int verbo; { struct section_offsets *section_offsets; asection *lowest_sect; struct cleanup *old_chain; init_entry_point_info (objfile); find_sym_fns (objfile); /* Make sure that partially constructed symbol tables will be cleaned up if an error occurs during symbol reading. */ old_chain = make_cleanup (free_objfile, objfile); if (mainline) { /* We will modify the main symbol table, make sure that all its users will be cleaned up if an error occurs during symbol reading. */ make_cleanup (clear_symtab_users, 0); /* Since no error yet, throw away the old symbol table. */ if (symfile_objfile != NULL) { free_objfile (symfile_objfile); symfile_objfile = NULL; } /* Currently we keep symbols from the add-symbol-file command. If the user wants to get rid of them, they should do "symbol-file" without arguments first. Not sure this is the best behavior (PR 2207). */ (*objfile -> sf -> sym_new_init) (objfile); } /* Convert addr into an offset rather than an absolute address. We find the lowest address of a loaded segment in the objfile, and assume that is where that got loaded. Due to historical precedent, we warn if that doesn't happen to be a text segment. */ if (mainline) { addr = 0; /* No offset from objfile addresses. */ } else { lowest_sect = bfd_get_section_by_name (objfile->obfd, ".text"); if (lowest_sect == NULL) bfd_map_over_sections (objfile->obfd, find_lowest_section, (PTR) &lowest_sect); if (lowest_sect == NULL) warning ("no loadable sections found in added symbol-file %s", objfile->name); else if ((bfd_get_section_flags (objfile->obfd, lowest_sect) & SEC_CODE) == 0) /* FIXME-32x64--assumes bfd_vma fits in long. */ warning ("Lowest section in %s is %s at 0x%lx", objfile->name, bfd_section_name (objfile->obfd, lowest_sect), (unsigned long) bfd_section_vma (objfile->obfd, lowest_sect)); if (lowest_sect) addr -= bfd_section_vma (objfile->obfd, lowest_sect); } /* Initialize symbol reading routines for this objfile, allow complaints to appear for this new file, and record how verbose to be, then do the initial symbol reading for this file. */ (*objfile -> sf -> sym_init) (objfile); clear_complaints (1, verbo); section_offsets = (*objfile -> sf -> sym_offsets) (objfile, addr); objfile->section_offsets = section_offsets; #ifndef IBM6000_TARGET /* This is a SVR4/SunOS specific hack, I think. In any event, it screws RS/6000. sym_offsets should be doing this sort of thing, because it knows the mapping between bfd sections and section_offsets. */ /* This is a hack. As far as I can tell, section offsets are not target dependent. They are all set to addr with a couple of exceptions. The exceptions are sysvr4 shared libraries, whose offsets are kept in solib structures anyway and rs6000 xcoff which handles shared libraries in a completely unique way. Section offsets are built similarly, except that they are built by adding addr in all cases because there is no clear mapping from section_offsets into actual sections. Note that solib.c has a different algorythm for finding section offsets. These should probably all be collapsed into some target independent form of shared library support. FIXME. */ if (addr) { struct obj_section *s; for (s = objfile->sections; s < objfile->sections_end; ++s) { s->addr -= s->offset; s->addr += addr; s->endaddr -= s->offset; s->endaddr += addr; s->offset += addr; } } #endif /* not IBM6000_TARGET */ (*objfile -> sf -> sym_read) (objfile, section_offsets, mainline); if (!have_partial_symbols () && !have_full_symbols ()) { wrap_here (""); printf_filtered ("(no debugging symbols found)..."); wrap_here (""); } /* Don't allow char * to have a typename (else would get caddr_t). Ditto void *. FIXME: Check whether this is now done by all the symbol readers themselves (many of them now do), and if so remove it from here. */ TYPE_NAME (lookup_pointer_type (builtin_type_char)) = 0; TYPE_NAME (lookup_pointer_type (builtin_type_void)) = 0; /* Mark the objfile has having had initial symbol read attempted. Note that this does not mean we found any symbols... */ objfile -> flags |= OBJF_SYMS; /* Discard cleanups as symbol reading was successful. */ discard_cleanups (old_chain); /* Call this after reading in a new symbol table to give target dependant code a crack at the new symbols. For instance, this could be used to update the values of target-specific symbols GDB needs to keep track of (such as _sigtramp, or whatever). */ TARGET_SYMFILE_POSTREAD (objfile); } /* Perform required actions after either reading in the initial symbols for a new objfile, or mapping in the symbols from a reusable objfile. */ void new_symfile_objfile (objfile, mainline, verbo) struct objfile *objfile; int mainline; int verbo; { /* If this is the main symbol file we have to clean up all users of the old main symbol file. Otherwise it is sufficient to fixup all the breakpoints that may have been redefined by this symbol file. */ if (mainline) { /* OK, make it the "real" symbol file. */ symfile_objfile = objfile; clear_symtab_users (); } else { breakpoint_re_set (); } /* We're done reading the symbol file; finish off complaints. */ clear_complaints (0, verbo); } /* Process a symbol file, as either the main file or as a dynamically loaded file. NAME is the file name (which will be tilde-expanded and made absolute herein) (but we don't free or modify NAME itself). FROM_TTY says how verbose to be. MAINLINE specifies whether this is the main symbol file, or whether it's an extra symbol file such as dynamically loaded code. If !mainline, ADDR is the address where the text segment was loaded. Upon success, returns a pointer to the objfile that was added. Upon failure, jumps back to command level (never returns). */ struct objfile * symbol_file_add (name, from_tty, addr, mainline, mapped, readnow) char *name; int from_tty; CORE_ADDR addr; int mainline; int mapped; int readnow; { struct objfile *objfile; struct partial_symtab *psymtab; bfd *abfd; /* Open a bfd for the file, and give user a chance to burp if we'd be interactively wiping out any existing symbols. */ abfd = symfile_bfd_open (name); if ((have_full_symbols () || have_partial_symbols ()) && mainline && from_tty && !query ("Load new symbol table from \"%s\"? ", name)) error ("Not confirmed."); objfile = allocate_objfile (abfd, mapped); /* If the objfile uses a mapped symbol file, and we have a psymtab for it, then skip reading any symbols at this time. */ if ((objfile -> flags & OBJF_MAPPED) && (objfile -> flags & OBJF_SYMS)) { /* We mapped in an existing symbol table file that already has had initial symbol reading performed, so we can skip that part. Notify the user that instead of reading the symbols, they have been mapped. */ if (from_tty || info_verbose) { printf_filtered ("Mapped symbols for %s...", name); wrap_here (""); gdb_flush (gdb_stdout); } init_entry_point_info (objfile); find_sym_fns (objfile); } else { /* We either created a new mapped symbol table, mapped an existing symbol table file which has not had initial symbol reading performed, or need to read an unmapped symbol table. */ if (from_tty || info_verbose) { printf_filtered ("Reading symbols from %s...", name); wrap_here (""); gdb_flush (gdb_stdout); } syms_from_objfile (objfile, addr, mainline, from_tty); } /* We now have at least a partial symbol table. Check to see if the user requested that all symbols be read on initial access via either the gdb startup command line or on a per symbol file basis. Expand all partial symbol tables for this objfile if so. */ if (readnow || readnow_symbol_files) { if (from_tty || info_verbose) { printf_filtered ("expanding to full symbols..."); wrap_here (""); gdb_flush (gdb_stdout); } for (psymtab = objfile -> psymtabs; psymtab != NULL; psymtab = psymtab -> next) { psymtab_to_symtab (psymtab); } } if (from_tty || info_verbose) { printf_filtered ("done.\n"); gdb_flush (gdb_stdout); } new_symfile_objfile (objfile, mainline, from_tty); return (objfile); } /* This is the symbol-file command. Read the file, analyze its symbols, and add a struct symtab to a symtab list. The syntax of the command is rather bizarre--(1) buildargv implements various quoting conventions which are undocumented and have little or nothing in common with the way things are quoted (or not quoted) elsewhere in GDB, (2) options are used, which are not generally used in GDB (perhaps "set mapped on", "set readnow on" would be better), (3) the order of options matters, which is contrary to GNU conventions (because it is confusing and inconvenient). */ void symbol_file_command (args, from_tty) char *args; int from_tty; { char **argv; char *name = NULL; CORE_ADDR text_relocation = 0; /* text_relocation */ struct cleanup *cleanups; int mapped = 0; int readnow = 0; dont_repeat (); if (args == NULL) { if ((have_full_symbols () || have_partial_symbols ()) && from_tty && !query ("Discard symbol table from `%s'? ", symfile_objfile -> name)) error ("Not confirmed."); free_all_objfiles (); symfile_objfile = NULL; if (from_tty) { printf_unfiltered ("No symbol file now.\n"); } } else { if ((argv = buildargv (args)) == NULL) { nomem (0); } cleanups = make_cleanup (freeargv, (char *) argv); while (*argv != NULL) { if (STREQ (*argv, "-mapped")) { mapped = 1; } else if (STREQ (*argv, "-readnow")) { readnow = 1; } else if (**argv == '-') { error ("unknown option `%s'", *argv); } else { char *p; name = *argv; /* this is for rombug remote only, to get the text relocation by using link command */ p = strrchr(name, '/'); if (p != NULL) p++; else p = name; target_link(p, &text_relocation); if (text_relocation == (CORE_ADDR)0) return; else if (text_relocation == (CORE_ADDR)-1) symbol_file_add (name, from_tty, (CORE_ADDR)0, 1, mapped, readnow); else symbol_file_add (name, from_tty, (CORE_ADDR)text_relocation, 0, mapped, readnow); /* Getting new symbols may change our opinion about what is frameless. */ reinit_frame_cache (); set_initial_language (); } argv++; } if (name == NULL) { error ("no symbol file name was specified"); } do_cleanups (cleanups); } } /* Set the initial language. A better solution would be to record the language in the psymtab when reading partial symbols, and then use it (if known) to set the language. This would be a win for formats that encode the language in an easily discoverable place, such as DWARF. For stabs, we can jump through hoops looking for specially named symbols or try to intuit the language from the specific type of stabs we find, but we can't do that until later when we read in full symbols. FIXME. */ static void set_initial_language () { struct partial_symtab *pst; enum language lang = language_unknown; pst = find_main_psymtab (); if (pst != NULL) { if (pst -> filename != NULL) { lang = deduce_language_from_filename (pst -> filename); } if (lang == language_unknown) { /* Make C the default language */ lang = language_c; } set_language (lang); expected_language = current_language; /* Don't warn the user */ } } /* Open file specified by NAME and hand it off to BFD for preliminary analysis. Result is a newly initialized bfd *, which includes a newly malloc'd` copy of NAME (tilde-expanded and made absolute). In case of trouble, error() is called. */ static bfd * symfile_bfd_open (name) char *name; { bfd *sym_bfd; int desc; char *absolute_name; name = tilde_expand (name); /* Returns 1st new malloc'd copy */ /* Look down path for it, allocate 2nd new malloc'd copy. */ desc = openp (getenv ("PATH"), 1, name, O_RDONLY | O_BINARY, 0, &absolute_name); if (desc < 0) { make_cleanup (free, name); perror_with_name (name); } free (name); /* Free 1st new malloc'd copy */ name = absolute_name; /* Keep 2nd malloc'd copy in bfd */ /* It'll be freed in free_objfile(). */ sym_bfd = bfd_fdopenr (name, gnutarget, desc); if (!sym_bfd) { close (desc); make_cleanup (free, name); error ("\"%s\": can't open to read symbols: %s.", name, bfd_errmsg (bfd_get_error ())); } sym_bfd->cacheable = true; if (!bfd_check_format (sym_bfd, bfd_object)) { /* FIXME: should be checking for errors from bfd_close (for one thing, on error it does not free all the storage associated with the bfd). */ bfd_close (sym_bfd); /* This also closes desc */ make_cleanup (free, name); error ("\"%s\": can't read symbols: %s.", name, bfd_errmsg (bfd_get_error ())); } return (sym_bfd); } /* Link a new symtab_fns into the global symtab_fns list. Called on gdb startup by the _initialize routine in each object file format reader, to register information about each format the the reader is prepared to handle. */ void add_symtab_fns (sf) struct sym_fns *sf; { sf->next = symtab_fns; symtab_fns = sf; } /* Initialize to read symbols from the symbol file sym_bfd. It either returns or calls error(). The result is an initialized struct sym_fns in the objfile structure, that contains cached information about the symbol file. */ static void find_sym_fns (objfile) struct objfile *objfile; { struct sym_fns *sf; enum bfd_flavour our_flavour = bfd_get_flavour (objfile -> obfd); char *our_target = bfd_get_target (objfile -> obfd); /* Special kludge for RS/6000. See xcoffread.c. */ if (STREQ (our_target, "aixcoff-rs6000")) our_flavour = (enum bfd_flavour)-1; /* Special kludge for apollo. See dstread.c. */ if (STREQN (our_target, "apollo", 6)) our_flavour = (enum bfd_flavour)-2; for (sf = symtab_fns; sf != NULL; sf = sf -> next) { if (our_flavour == sf -> sym_flavour) { objfile -> sf = sf; return; } } error ("I'm sorry, Dave, I can't do that. Symbol format `%s' unknown.", bfd_get_target (objfile -> obfd)); } /* This function runs the load command of our current target. */ static void load_command (arg, from_tty) char *arg; int from_tty; { if (arg == NULL) arg = get_exec_file (1); target_load (arg, from_tty); } /* This version of "load" should be usable for any target. Currently it is just used for remote targets, not inftarg.c or core files, on the theory that only in that case is it useful. Avoiding xmodem and the like seems like a win (a) because we don't have to worry about finding it, and (b) On VMS, fork() is very slow and so we don't want to run a subprocess. On the other hand, I'm not sure how performance compares. */ void generic_load (filename, from_tty) char *filename; int from_tty; { struct cleanup *old_cleanups; asection *s; bfd *loadfile_bfd; loadfile_bfd = bfd_openr (filename, gnutarget); if (loadfile_bfd == NULL) { perror_with_name (filename); return; } /* FIXME: should be checking for errors from bfd_close (for one thing, on error it does not free all the storage associated with the bfd). */ old_cleanups = make_cleanup (bfd_close, loadfile_bfd); if (!bfd_check_format (loadfile_bfd, bfd_object)) { error ("\"%s\" is not an object file: %s", filename, bfd_errmsg (bfd_get_error ())); } for (s = loadfile_bfd->sections; s; s = s->next) { if (s->flags & SEC_LOAD) { bfd_size_type size; size = bfd_get_section_size_before_reloc (s); if (size > 0) { char *buffer; struct cleanup *old_chain; bfd_vma vma; buffer = xmalloc (size); old_chain = make_cleanup (free, buffer); vma = bfd_get_section_vma (loadfile_bfd, s); /* Is this really necessary? I guess it gives the user something to look at during a long download. */ printf_filtered ("Loading section %s, size 0x%lx vma ", bfd_get_section_name (loadfile_bfd, s), (unsigned long) size); print_address_numeric (vma, 1, gdb_stdout); printf_filtered ("\n"); bfd_get_section_contents (loadfile_bfd, s, buffer, 0, size); target_write_memory (vma, buffer, size); do_cleanups (old_chain); } } } /* We were doing this in remote-mips.c, I suspect it is right for other targets too. */ write_pc (loadfile_bfd->start_address); /* FIXME: are we supposed to call symbol_file_add or not? According to a comment from remote-mips.c (where a call to symbol_file_add was commented out), making the call confuses GDB if more than one file is loaded in. remote-nindy.c had no call to symbol_file_add, but remote-vx.c does. */ do_cleanups (old_cleanups); } /* This function allows the addition of incrementally linked object files. It does not modify any state in the target, only in the debugger. */ /* ARGSUSED */ static void add_symbol_file_command (args, from_tty) char *args; int from_tty; { char *name = NULL; CORE_ADDR text_addr; char *arg; int readnow = 0; int mapped = 0; dont_repeat (); if (args == NULL) { error ("add-symbol-file takes a file name and an address"); } /* Make a copy of the string that we can safely write into. */ args = strdup (args); make_cleanup (free, args); /* Pick off any -option args and the file name. */ while ((*args != '\000') && (name == NULL)) { while (isspace (*args)) {args++;} arg = args; while ((*args != '\000') && !isspace (*args)) {args++;} if (*args != '\000') { *args++ = '\000'; } if (*arg != '-') { name = arg; } else if (STREQ (arg, "-mapped")) { mapped = 1; } else if (STREQ (arg, "-readnow")) { readnow = 1; } else { error ("unknown option `%s'", arg); } } /* After picking off any options and the file name, args should be left pointing at the remainder of the command line, which should be the address expression to evaluate. */ if (name == NULL) { error ("add-symbol-file takes a file name"); } name = tilde_expand (name); make_cleanup (free, name); if (*args != '\000') { text_addr = parse_and_eval_address (args); } else { target_link(name, &text_addr); if (text_addr == (CORE_ADDR)-1) error("Don't know how to get text start location for this file"); } /* FIXME-32x64: Assumes text_addr fits in a long. */ if (!query ("add symbol table from file \"%s\" at text_addr = %s?\n", name, local_hex_string ((unsigned long)text_addr))) error ("Not confirmed."); symbol_file_add (name, 0, text_addr, 0, mapped, readnow); /* Getting new symbols may change our opinion about what is frameless. */ reinit_frame_cache (); } static void add_shared_symbol_files_command (args, from_tty) char *args; int from_tty; { #ifdef ADD_SHARED_SYMBOL_FILES ADD_SHARED_SYMBOL_FILES (args, from_tty); #else error ("This command is not available in this configuration of GDB."); #endif } /* Re-read symbols if a symbol-file has changed. */ void reread_symbols () { struct objfile *objfile; long new_modtime; int reread_one = 0; struct stat new_statbuf; int res; /* With the addition of shared libraries, this should be modified, the load time should be saved in the partial symbol tables, since different tables may come from different source files. FIXME. This routine should then walk down each partial symbol table and see if the symbol table that it originates from has been changed */ for (objfile = object_files; objfile; objfile = objfile->next) { if (objfile->obfd) { #ifdef IBM6000_TARGET /* If this object is from a shared library, then you should stat on the library name, not member name. */ if (objfile->obfd->my_archive) res = stat (objfile->obfd->my_archive->filename, &new_statbuf); else #endif res = stat (objfile->name, &new_statbuf); if (res != 0) { /* FIXME, should use print_sys_errmsg but it's not filtered. */ printf_filtered ("`%s' has disappeared; keeping its symbols.\n", objfile->name); continue; } new_modtime = new_statbuf.st_mtime; if (new_modtime != objfile->mtime) { struct cleanup *old_cleanups; struct section_offsets *offsets; int num_offsets; int section_offsets_size; char *obfd_filename; printf_filtered ("`%s' has changed; re-reading symbols.\n", objfile->name); /* There are various functions like symbol_file_add, symfile_bfd_open, syms_from_objfile, etc., which might appear to do what we want. But they have various other effects which we *don't* want. So we just do stuff ourselves. We don't worry about mapped files (for one thing, any mapped file will be out of date). */ /* If we get an error, blow away this objfile (not sure if that is the correct response for things like shared libraries). */ old_cleanups = make_cleanup (free_objfile, objfile); /* We need to do this whenever any symbols go away. */ make_cleanup (clear_symtab_users, 0); /* Clean up any state BFD has sitting around. We don't need to close the descriptor but BFD lacks a way of closing the BFD without closing the descriptor. */ obfd_filename = bfd_get_filename (objfile->obfd); if (!bfd_close (objfile->obfd)) error ("Can't close BFD for %s: %s", objfile->name, bfd_errmsg (bfd_get_error ())); objfile->obfd = bfd_openr (obfd_filename, gnutarget); if (objfile->obfd == NULL) error ("Can't open %s to read symbols.", objfile->name); /* bfd_openr sets cacheable to true, which is what we want. */ if (!bfd_check_format (objfile->obfd, bfd_object)) error ("Can't read symbols from %s: %s.", objfile->name, bfd_errmsg (bfd_get_error ())); /* Save the offsets, we will nuke them with the rest of the psymbol_obstack. */ num_offsets = objfile->num_sections; section_offsets_size = sizeof (struct section_offsets) + sizeof (objfile->section_offsets->offsets) * num_offsets; offsets = (struct section_offsets *) alloca (section_offsets_size); memcpy (offsets, objfile->section_offsets, section_offsets_size); /* Nuke all the state that we will re-read. Much of the following code which sets things to NULL really is necessary to tell other parts of GDB that there is nothing currently there. */ /* FIXME: Do we have to free a whole linked list, or is this enough? */ if (objfile->global_psymbols.list) mfree (objfile->md, objfile->global_psymbols.list); objfile->global_psymbols.list = NULL; objfile->global_psymbols.next = NULL; objfile->global_psymbols.size = 0; if (objfile->static_psymbols.list) mfree (objfile->md, objfile->static_psymbols.list); objfile->static_psymbols.list = NULL; objfile->static_psymbols.next = NULL; objfile->static_psymbols.size = 0; /* Free the obstacks for non-reusable objfiles */ obstack_free (&objfile -> psymbol_obstack, 0); obstack_free (&objfile -> symbol_obstack, 0); obstack_free (&objfile -> type_obstack, 0); objfile->sections = NULL; objfile->symtabs = NULL; objfile->psymtabs = NULL; objfile->free_psymtabs = NULL; objfile->msymbols = NULL; objfile->minimal_symbol_count= 0; objfile->fundamental_types = NULL; if (objfile -> sf != NULL) { (*objfile -> sf -> sym_finish) (objfile); } /* We never make this a mapped file. */ objfile -> md = NULL; /* obstack_specify_allocation also initializes the obstack so it is empty. */ obstack_specify_allocation (&objfile -> psymbol_obstack, 0, 0, xmalloc, free); obstack_specify_allocation (&objfile -> symbol_obstack, 0, 0, xmalloc, free); obstack_specify_allocation (&objfile -> type_obstack, 0, 0, xmalloc, free); if (build_objfile_section_table (objfile)) { error ("Can't find the file sections in `%s': %s", objfile -> name, bfd_errmsg (bfd_get_error ())); } /* We use the same section offsets as from last time. I'm not sure whether that is always correct for shared libraries. */ objfile->section_offsets = (struct section_offsets *) obstack_alloc (&objfile -> psymbol_obstack, section_offsets_size); memcpy (objfile->section_offsets, offsets, section_offsets_size); objfile->num_sections = num_offsets; /* What the hell is sym_new_init for, anyway? The concept of distinguishing between the main file and additional files in this way seems rather dubious. */ if (objfile == symfile_objfile) (*objfile->sf->sym_new_init) (objfile); (*objfile->sf->sym_init) (objfile); clear_complaints (1, 1); /* The "mainline" parameter is a hideous hack; I think leaving it zero is OK since dbxread.c also does what it needs to do if objfile->global_psymbols.size is 0. */ (*objfile->sf->sym_read) (objfile, objfile->section_offsets, 0); if (!have_partial_symbols () && !have_full_symbols ()) { wrap_here (""); printf_filtered ("(no debugging symbols found)\n"); wrap_here (""); } objfile -> flags |= OBJF_SYMS; /* We're done reading the symbol file; finish off complaints. */ clear_complaints (0, 1); /* Getting new symbols may change our opinion about what is frameless. */ reinit_frame_cache (); /* Discard cleanups as symbol reading was successful. */ discard_cleanups (old_cleanups); /* If the mtime has changed between the time we set new_modtime and now, we *want* this to be out of date, so don't call stat again now. */ objfile->mtime = new_modtime; reread_one = 1; /* Call this after reading in a new symbol table to give target dependant code a crack at the new symbols. For instance, this could be used to update the values of target-specific symbols GDB needs to keep track of (such as _sigtramp, or whatever). */ TARGET_SYMFILE_POSTREAD (objfile); } } } if (reread_one) clear_symtab_users (); } enum language deduce_language_from_filename (filename) char *filename; { char *c; if (0 == filename) ; /* Get default */ else if (0 == (c = strrchr (filename, '.'))) ; /* Get default. */ else if (STREQ (c, ".c")) return language_c; else if (STREQ (c, ".cc") || STREQ (c, ".C") || STREQ (c, ".cxx") || STREQ (c, ".cpp") || STREQ (c, ".cp") || STREQ (c, ".c++")) return language_cplus; else if (STREQ (c, ".ch") || STREQ (c, ".c186") || STREQ (c, ".c286")) return language_chill; else if (STREQ (c, ".f") || STREQ (c, ".F")) return language_fortran; else if (STREQ (c, ".mod")) return language_m2; else if (STREQ (c, ".s") || STREQ (c, ".S")) return language_asm; return language_unknown; /* default */ } /* allocate_symtab: Allocate and partly initialize a new symbol table. Return a pointer to it. error() if no space. Caller must set these fields: LINETABLE(symtab) symtab->blockvector symtab->dirname symtab->free_code symtab->free_ptr initialize any EXTRA_SYMTAB_INFO possibly free_named_symtabs (symtab->filename); */ struct symtab * allocate_symtab (filename, objfile) char *filename; struct objfile *objfile; { register struct symtab *symtab; symtab = (struct symtab *) obstack_alloc (&objfile -> symbol_obstack, sizeof (struct symtab)); memset (symtab, 0, sizeof (*symtab)); symtab -> filename = obsavestring (filename, strlen (filename), &objfile -> symbol_obstack); symtab -> fullname = NULL; symtab -> language = deduce_language_from_filename (filename); /* Hook it to the objfile it comes from */ symtab -> objfile = objfile; symtab -> next = objfile -> symtabs; objfile -> symtabs = symtab; #ifdef INIT_EXTRA_SYMTAB_INFO INIT_EXTRA_SYMTAB_INFO (symtab); #endif return (symtab); } struct partial_symtab * allocate_psymtab (filename, objfile) char *filename; struct objfile *objfile; { struct partial_symtab *psymtab; if (objfile -> free_psymtabs) { psymtab = objfile -> free_psymtabs; objfile -> free_psymtabs = psymtab -> next; } else psymtab = (struct partial_symtab *) obstack_alloc (&objfile -> psymbol_obstack, sizeof (struct partial_symtab)); memset (psymtab, 0, sizeof (struct partial_symtab)); psymtab -> filename = obsavestring (filename, strlen (filename), &objfile -> psymbol_obstack); psymtab -> symtab = NULL; /* Hook it to the objfile it comes from */ psymtab -> objfile = objfile; psymtab -> next = objfile -> psymtabs; objfile -> psymtabs = psymtab; return (psymtab); } /* Reset all data structures in gdb which may contain references to symbol table date. */ void clear_symtab_users () { /* Someday, we should do better than this, by only blowing away the things that really need to be blown. */ clear_value_history (); clear_displays (); clear_internalvars (); breakpoint_re_set (); set_default_breakpoint (0, 0, 0, 0); current_source_symtab = 0; current_source_line = 0; clear_pc_function_cache (); } /* clear_symtab_users_once: This function is run after symbol reading, or from a cleanup. If an old symbol table was obsoleted, the old symbol table has been blown away, but the other GDB data structures that may reference it have not yet been cleared or re-directed. (The old symtab was zapped, and the cleanup queued, in free_named_symtab() below.) This function can be queued N times as a cleanup, or called directly; it will do all the work the first time, and then will be a no-op until the next time it is queued. This works by bumping a counter at queueing time. Much later when the cleanup is run, or at the end of symbol processing (in case the cleanup is discarded), if the queued count is greater than the "done-count", we do the work and set the done-count to the queued count. If the queued count is less than or equal to the done-count, we just ignore the call. This is needed because reading a single .o file will often replace many symtabs (one per .h file, for example), and we don't want to reset the breakpoints N times in the user's face. The reason we both queue a cleanup, and call it directly after symbol reading, is because the cleanup protects us in case of errors, but is discarded if symbol reading is successful. */ #if 0 /* FIXME: As free_named_symtabs is currently a big noop this function is no longer needed. */ static void clear_symtab_users_once PARAMS ((void)); static int clear_symtab_users_queued; static int clear_symtab_users_done; static void clear_symtab_users_once () { /* Enforce once-per-`do_cleanups'-semantics */ if (clear_symtab_users_queued <= clear_symtab_users_done) return; clear_symtab_users_done = clear_symtab_users_queued; clear_symtab_users (); } #endif /* Delete the specified psymtab, and any others that reference it. */ static void cashier_psymtab (pst) struct partial_symtab *pst; { struct partial_symtab *ps, *pprev = NULL; int i; /* Find its previous psymtab in the chain */ for (ps = pst->objfile->psymtabs; ps; ps = ps->next) { if (ps == pst) break; pprev = ps; } if (ps) { /* Unhook it from the chain. */ if (ps == pst->objfile->psymtabs) pst->objfile->psymtabs = ps->next; else pprev->next = ps->next; /* FIXME, we can't conveniently deallocate the entries in the partial_symbol lists (global_psymbols/static_psymbols) that this psymtab points to. These just take up space until all the psymtabs are reclaimed. Ditto the dependencies list and filename, which are all in the psymbol_obstack. */ /* We need to cashier any psymtab that has this one as a dependency... */ again: for (ps = pst->objfile->psymtabs; ps; ps = ps->next) { for (i = 0; i < ps->number_of_dependencies; i++) { if (ps->dependencies[i] == pst) { cashier_psymtab (ps); goto again; /* Must restart, chain has been munged. */ } } } } } /* If a symtab or psymtab for filename NAME is found, free it along with any dependent breakpoints, displays, etc. Used when loading new versions of object modules with the "add-file" command. This is only called on the top-level symtab or psymtab's name; it is not called for subsidiary files such as .h files. Return value is 1 if we blew away the environment, 0 if not. FIXME. The return valu appears to never be used. FIXME. I think this is not the best way to do this. We should work on being gentler to the environment while still cleaning up all stray pointers into the freed symtab. */ int free_named_symtabs (name) char *name; { #if 0 /* FIXME: With the new method of each objfile having it's own psymtab list, this function needs serious rethinking. In particular, why was it ever necessary to toss psymtabs with specific compilation unit filenames, as opposed to all psymtabs from a particular symbol file? -- fnf Well, the answer is that some systems permit reloading of particular compilation units. We want to blow away any old info about these compilation units, regardless of which objfiles they arrived in. --gnu. */ register struct symtab *s; register struct symtab *prev; register struct partial_symtab *ps; struct blockvector *bv; int blewit = 0; /* We only wack things if the symbol-reload switch is set. */ if (!symbol_reloading) return 0; /* Some symbol formats have trouble providing file names... */ if (name == 0 || *name == '\0') return 0; /* Look for a psymtab with the specified name. */ again2: for (ps = partial_symtab_list; ps; ps = ps->next) { if (STREQ (name, ps->filename)) { cashier_psymtab (ps); /* Blow it away...and its little dog, too. */ goto again2; /* Must restart, chain has been munged */ } } /* Look for a symtab with the specified name. */ for (s = symtab_list; s; s = s->next) { if (STREQ (name, s->filename)) break; prev = s; } if (s) { if (s == symtab_list) symtab_list = s->next; else prev->next = s->next; /* For now, queue a delete for all breakpoints, displays, etc., whether or not they depend on the symtab being freed. This should be changed so that only those data structures affected are deleted. */ /* But don't delete anything if the symtab is empty. This test is necessary due to a bug in "dbxread.c" that causes empty symtabs to be created for N_SO symbols that contain the pathname of the object file. (This problem has been fixed in GDB 3.9x). */ bv = BLOCKVECTOR (s); if (BLOCKVECTOR_NBLOCKS (bv) > 2 || BLOCK_NSYMS (BLOCKVECTOR_BLOCK (bv, GLOBAL_BLOCK)) || BLOCK_NSYMS (BLOCKVECTOR_BLOCK (bv, STATIC_BLOCK))) { complain (&oldsyms_complaint, name); clear_symtab_users_queued++; make_cleanup (clear_symtab_users_once, 0); blewit = 1; } else { complain (&empty_symtab_complaint, name); } free_symtab (s); } else { /* It is still possible that some breakpoints will be affected even though no symtab was found, since the file might have been compiled without debugging, and hence not be associated with a symtab. In order to handle this correctly, we would need to keep a list of text address ranges for undebuggable files. For now, we do nothing, since this is a fairly obscure case. */ ; } /* FIXME, what about the minimal symbol table? */ return blewit; #else return (0); #endif } /* Allocate and partially fill a partial symtab. It will be completely filled at the end of the symbol list. SYMFILE_NAME is the name of the symbol-file we are reading from, and ADDR is the address relative to which its symbols are (incremental) or 0 (normal). */ struct partial_symtab * start_psymtab_common (objfile, section_offsets, filename, textlow, global_syms, static_syms) struct objfile *objfile; struct section_offsets *section_offsets; char *filename; CORE_ADDR textlow; struct partial_symbol *global_syms; struct partial_symbol *static_syms; { struct partial_symtab *psymtab; psymtab = allocate_psymtab (filename, objfile); psymtab -> section_offsets = section_offsets; psymtab -> textlow = textlow; psymtab -> texthigh = psymtab -> textlow; /* default */ psymtab -> globals_offset = global_syms - objfile -> global_psymbols.list; psymtab -> statics_offset = static_syms - objfile -> static_psymbols.list; return (psymtab); } /* Debugging versions of functions that are usually inline macros (see symfile.h). */ #if !INLINE_ADD_PSYMBOL /* Add a symbol with a long value to a psymtab. Since one arg is a struct, we pass in a ptr and deref it (sigh). */ void add_psymbol_to_list (name, namelength, namespace, class, list, val, language, objfile) char *name; int namelength; enum namespace namespace; enum address_class class; struct psymbol_allocation_list *list; long val; enum language language; struct objfile *objfile; { register struct partial_symbol *psym; register char *demangled_name; if (list->next >= list->list + list->size) { extend_psymbol_list (list,objfile); } psym = list->next++; SYMBOL_NAME (psym) = (char *) obstack_alloc (&objfile->psymbol_obstack, namelength + 1); memcpy (SYMBOL_NAME (psym), name, namelength); SYMBOL_NAME (psym)[namelength] = '\0'; SYMBOL_VALUE (psym) = val; SYMBOL_SECTION (psym) = 0; SYMBOL_LANGUAGE (psym) = language; PSYMBOL_NAMESPACE (psym) = namespace; PSYMBOL_CLASS (psym) = class; SYMBOL_INIT_LANGUAGE_SPECIFIC (psym, language); OBJSTAT (objfile, psyms++); } /* Add a symbol with a CORE_ADDR value to a psymtab. */ void add_psymbol_addr_to_list (name, namelength, namespace, class, list, val, language, objfile) char *name; int namelength; enum namespace namespace; enum address_class class; struct psymbol_allocation_list *list; CORE_ADDR val; enum language language; struct objfile *objfile; { register struct partial_symbol *psym; register char *demangled_name; if (list->next >= list->list + list->size) { extend_psymbol_list (list,objfile); } psym = list->next++; SYMBOL_NAME (psym) = (char *) obstack_alloc (&objfile->psymbol_obstack, namelength + 1); memcpy (SYMBOL_NAME (psym), name, namelength); SYMBOL_NAME (psym)[namelength] = '\0'; SYMBOL_VALUE_ADDRESS (psym) = val; SYMBOL_SECTION (psym) = 0; SYMBOL_LANGUAGE (psym) = language; PSYMBOL_NAMESPACE (psym) = namespace; PSYMBOL_CLASS (psym) = class; SYMBOL_INIT_LANGUAGE_SPECIFIC (psym, language); OBJSTAT (objfile, psyms++); } #endif /* !INLINE_ADD_PSYMBOL */ /* Initialize storage for partial symbols. */ void init_psymbol_list (objfile, total_symbols) struct objfile *objfile; int total_symbols; { /* Free any previously allocated psymbol lists. */ if (objfile -> global_psymbols.list) { mfree (objfile -> md, (PTR)objfile -> global_psymbols.list); } if (objfile -> static_psymbols.list) { mfree (objfile -> md, (PTR)objfile -> static_psymbols.list); } /* Current best guess is that approximately a twentieth of the total symbols (in a debugging file) are global or static oriented symbols */ objfile -> global_psymbols.size = total_symbols / 10; objfile -> static_psymbols.size = total_symbols / 10; objfile -> global_psymbols.next = objfile -> global_psymbols.list = (struct partial_symbol *) xmmalloc (objfile -> md, objfile -> global_psymbols.size * sizeof (struct partial_symbol)); objfile -> static_psymbols.next = objfile -> static_psymbols.list = (struct partial_symbol *) xmmalloc (objfile -> md, objfile -> static_psymbols.size * sizeof (struct partial_symbol)); } void _initialize_symfile () { struct cmd_list_element *c; c = add_cmd ("symbol-file", class_files, symbol_file_command, "Load symbol table from executable file FILE.\n\ The `file' command can also load symbol tables, as well as setting the file\n\ to execute.", &cmdlist); c->completer = filename_completer; c = add_cmd ("add-symbol-file", class_files, add_symbol_file_command, "Usage: add-symbol-file FILE ADDR\n\ Load the symbols from FILE, assuming FILE has been dynamically loaded.\n\ ADDR is the starting address of the file's text.", &cmdlist); c->completer = filename_completer; c = add_cmd ("add-shared-symbol-files", class_files, add_shared_symbol_files_command, "Load the symbols from shared objects in the dynamic linker's link map.", &cmdlist); c = add_alias_cmd ("assf", "add-shared-symbol-files", class_files, 1, &cmdlist); c = add_cmd ("load", class_files, load_command, "Dynamically load FILE into the running program, and record its symbols\n\ for access from GDB.", &cmdlist); c->completer = filename_completer; add_show_from_set (add_set_cmd ("symbol-reloading", class_support, var_boolean, (char *)&symbol_reloading, "Set dynamic symbol table reloading multiple times in one run.", &setlist), &showlist); }