Mercurial > hg > xemacs-beta
view src/unexaix.c @ 4690:257b468bf2ca
Move the #'query-coding-region implementation to C.
This is necessary because there is no reasonable way to access the
corresponding mswindows-multibyte functionality from Lisp, and we need such
functionality if we're going to have a reliable and portable
#'query-coding-region implementation. However, this change doesn't yet
provide #'query-coding-region for the mswindow-multibyte coding systems,
there should be no functional differences between an XEmacs with this change
and one without it.
src/ChangeLog addition:
2009-09-19 Aidan Kehoe <kehoea@parhasard.net>
Move the #'query-coding-region implementation to C.
This is necessary because there is no reasonable way to access the
corresponding mswindows-multibyte functionality from Lisp, and we
need such functionality if we're going to have a reliable and
portable #'query-coding-region implementation. However, this
change doesn't yet provide #'query-coding-region for the
mswindow-multibyte coding systems, there should be no functional
differences between an XEmacs with this change and one without it.
* mule-coding.c (struct fixed_width_coding_system):
Add a new coding system type, fixed_width, and implement it. It
uses the CCL infrastructure but has a much simpler creation API,
and its own query_method, formerly in lisp/mule/mule-coding.el.
* unicode.c:
Move the Unicode query method implementation here from
unicode.el.
* lisp.h: Declare Fmake_coding_system_internal, Fcopy_range_table
here.
* intl-win32.c (complex_vars_of_intl_win32):
Use Fmake_coding_system_internal, not Fmake_coding_system.
* general-slots.h: Add Qsucceeded, Qunencodable, Qinvalid_sequence
here.
* file-coding.h (enum coding_system_variant):
Add fixed_width_coding_system here.
(struct coding_system_methods):
Add query_method and query_lstream_method to the coding system
methods.
Provide flags for the query methods.
Declare the default query method; initialise it correctly in
INITIALIZE_CODING_SYSTEM_TYPE.
* file-coding.c (default_query_method):
New function, the default query method for coding systems that do
not set it. Moved from coding.el.
(make_coding_system_1):
Accept new elements in PROPS in #'make-coding-system; aliases, a
list of aliases; safe-chars and safe-charsets (these were
previously accepted but not saved); and category.
(Fmake_coding_system_internal):
New function, what used to be #'make-coding-system--on Mule
builds, we've now moved some of the functionality of this to
Lisp.
(Fcoding_system_canonical_name_p):
Move this earlier in the file, since it's now called from within
make_coding_system_1.
(Fquery_coding_region):
Move the implementation of this here, from coding.el.
(complex_vars_of_file_coding):
Call Fmake_coding_system_internal, not Fmake_coding_system;
specify safe-charsets properties when we're a mule build.
* extents.h (mouse_highlight_priority, Fset_extent_priority,
Fset_extent_face, Fmap_extents):
Make these available to other C files.
lisp/ChangeLog addition:
2009-09-19 Aidan Kehoe <kehoea@parhasard.net>
Move the #'query-coding-region implementation to C.
* coding.el:
Consolidate code that depends on the presence or absence of Mule
at the end of this file.
(default-query-coding-region, query-coding-region):
Move these functions to C.
(default-query-coding-region-safe-charset-skip-chars-map):
Remove this variable, the corresponding C variable is
Vdefault_query_coding_region_chartab_cache in file-coding.c.
(query-coding-string): Update docstring to reflect actual multiple
values, be more careful about not modifying a range table that
we're currently mapping over.
(encode-coding-char): Make the implementation of this simpler.
(featurep 'mule): Autoload #'make-coding-system from
mule/make-coding-system.el if we're a mule build; provide an
appropriate compiler macro.
Do various non-mule compatibility things if we're not a mule
build.
* update-elc.el (additional-dump-dependencies):
Add mule/make-coding-system as a dump time dependency if we're a
mule build.
* unicode.el (ccl-encode-to-ucs-2):
(decode-char):
(encode-char):
Move these earlier in the file, for the sake of some byte compile
warnings.
(unicode-query-coding-region):
Move this to unicode.c
* mule/make-coding-system.el:
New file, not dumped. Contains the functionality to rework the
arguments necessary for fixed-width coding systems, and contains
the implementation of #'make-coding-system, which now calls
#'make-coding-system-internal.
* mule/vietnamese.el (viscii):
* mule/latin.el (iso-8859-2):
(windows-1250):
(iso-8859-3):
(iso-8859-4):
(iso-8859-14):
(iso-8859-15):
(iso-8859-16):
(iso-8859-9):
(macintosh):
(windows-1252):
* mule/hebrew.el (iso-8859-8):
* mule/greek.el (iso-8859-7):
(windows-1253):
* mule/cyrillic.el (iso-8859-5):
(koi8-r):
(koi8-u):
(windows-1251):
(alternativnyj):
(koi8-ru):
(koi8-t):
(koi8-c):
(koi8-o):
* mule/arabic.el (iso-8859-6):
(windows-1256):
Move all these coding systems to being of type fixed-width, not of
type CCL. This allows the distinct query-coding-region for them to
be in C, something which will eventually allow us to implement
query-coding-region for the mswindows-multibyte coding systems.
* mule/general-late.el (posix-charset-to-coding-system-hash):
Document why we're pre-emptively persuading the byte compiler that
the ELC for this file needs to be written using escape-quoted.
Call #'set-unicode-query-skip-chars-args, now the Unicode
query-coding-region implementation is in C.
* mule/thai-xtis.el (tis-620):
Don't bother checking whether we're XEmacs or not here.
* mule/mule-coding.el:
Move the eight bit fixed-width functionality from this file to
make-coding-system.el.
tests/ChangeLog addition:
2009-09-19 Aidan Kehoe <kehoea@parhasard.net>
* automated/mule-tests.el:
Check a coding system's type, not an 8-bit-fixed property, for
whether that coding system should be treated as a fixed-width
coding system.
* automated/query-coding-tests.el:
Don't test the query coding functionality for mswindows-multibyte
coding systems, it's not yet implemented.
author | Aidan Kehoe <kehoea@parhasard.net> |
---|---|
date | Sat, 19 Sep 2009 22:53:13 +0100 |
parents | facf3239ba30 |
children | aa5ed11f473b |
line wrap: on
line source
/* Dump an executable image. Copyright (C) 1985, 1986, 1987, 1988 Free Software Foundation, Inc. This file is part of XEmacs. XEmacs 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, or (at your option) any later version. XEmacs 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 XEmacs; see the file COPYING. If not, write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */ /* Synched up with: FSF 20.2. */ /* This file has been ... uhhhhh ... Mule-ized. Yeah. (Everything here is external format. That's potentially dangerous, but in practice it'll be OK.) --ben */ /* Originally based on the COFF unexec.c by Spencer W. Thomas. * * Subsequently hacked on by * Bill Mann <Bill_Man@praxisint.com> * Andrew Vignaux <Andrew.Vignaux@comp.vuw.ac.nz> * Mike Sperber <sperber@informatik.uni-tuebingen.de> * * Synopsis: * unexec (new_name, a_name, data_start, bss_start, entry_address) * char *new_name, *a_name; * unsigned data_start, bss_start, entry_address; * * Takes a snapshot of the program and makes an a.out format file in the * file named by the string argument new_name. * If a_name is non-NULL, the symbol table will be taken from the given file. * On some machines, an existing a_name file is required. * * data_start and entry_address are ignored. * * bss_start indicates how much of the data segment is to be saved in the * a.out file and restored when the program is executed. It gives the lowest * unsaved address, and is rounded up to a page boundary. The default when 0 * is given assumes that the entire data segment is to be stored, including * the previous data and bss as well as any additional storage allocated with * sbrk(2). * */ #ifndef emacs #define PERROR(arg) perror (arg); return -1 #else #include <config.h> #define PERROR(file) report_error (file, new_) #endif #include <a.out.h> /* Define getpagesize () if the system does not. Note that this may depend on symbols defined in a.out.h */ #include "getpagesize.h" #include <sys/types.h> #include <stdio.h> #include <sys/stat.h> #include <errno.h> #include <unistd.h> #include <fcntl.h> extern char *start_of_text (void); /* Start of text */ extern char *start_of_data (void); /* Start of initialized data */ extern int _data; extern int _text; #include <filehdr.h> #include <aouthdr.h> #include <scnhdr.h> #include <syms.h> static struct filehdr f_hdr; /* File header */ static struct aouthdr f_ohdr; /* Optional file header (a.out) */ static long bias; /* Bias to add for growth */ static long lnnoptr; /* Pointer to line-number info within file */ static long text_scnptr; static long data_scnptr; #define ALIGN(val, pwr) (((val) + ((1L<<(pwr))-1)) & ~((1L<<(pwr))-1)) static long load_scnptr; static long orig_load_scnptr; static long orig_data_scnptr; static int unrelocate_symbols (int, int, char *, char *); #ifndef MAX_SECTIONS #define MAX_SECTIONS 10 #endif static int adjust_lnnoptrs (int, int, char *); static int pagemask; /* Correct an int which is the bit pattern of a pointer to a byte into an int which is the number of a byte. This is a no-op on ordinary machines, but not on all. */ #ifndef ADDR_CORRECT /* Let m-*.h files override this definition */ #define ADDR_CORRECT(x) ((char *)(x) - (char*)0) #endif #ifdef emacs #include "lisp.h" static void report_error (char *file, int fd) { if (fd) close (fd); report_error_with_errno (Qio_error, "Cannot unexec", build_string (file)); } #endif /* emacs */ #define ERROR0(msg) report_error_1 (new_, msg, 0, 0); return -1 #define ERROR1(msg,x) report_error_1 (new_, msg, x, 0); return -1 #define ERROR2(msg,x,y) report_error_1 (new_, msg, x, y); return -1 static void report_error_1 (int fd, char *msg, int a1, int a2) { close (fd); #ifdef emacs signal_ferror (Qio_error, msg, a1, a2); #else fprintf (stderr, msg, a1, a2); fprintf (stderr, "\n"); #endif } static int make_hdr (int, int, unsigned, unsigned, unsigned, char *, char *); static void mark_x (char *); static int copy_text_and_data (int); static int copy_sym (int, int, char *, char *); static void write_segment (int, char *, char *); /* **************************************************************** * unexec * * driving logic. */ int unexec (char *new_name, char *a_name, uintptr_t data_start, uintptr_t bss_start, uintptr_t entry_address) { int new_ = -1, a_out = -1; if (a_name && (a_out = open (a_name, O_RDONLY)) < 0) { PERROR (a_name); } if ((new_ = creat (new_name, 0666)) < 0) { PERROR (new_name); } if (make_hdr (new_, a_out, data_start, bss_start, entry_address, a_name, new_name) < 0 || copy_text_and_data (new_) < 0 || copy_sym (new_, a_out, a_name, new_name) < 0 || adjust_lnnoptrs (new_, a_out, new_name) < 0 || unrelocate_symbols (new_, a_out, a_name, new_name) < 0) { close (new_); return -1; } close (new_); if (a_out >= 0) close (a_out); mark_x (new_name); return 0; } /* **************************************************************** * make_hdr * * Make the header in the new a.out from the header in core. * Modify the text and data sizes. */ static int make_hdr (int new_, int a_out, unsigned data_start, unsigned bss_start, unsigned UNUSED (entry_address), char *a_name, char *new_name) { int scns; unsigned int bss_end; struct scnhdr section[MAX_SECTIONS]; struct scnhdr * f_thdr; /* Text section header */ struct scnhdr * f_dhdr; /* Data section header */ struct scnhdr * f_bhdr; /* Bss section header */ struct scnhdr * f_lhdr; /* Loader section header */ struct scnhdr * f_tchdr; /* Typechk section header */ struct scnhdr * f_dbhdr; /* Debug section header */ struct scnhdr * f_xhdr; /* Except section header */ load_scnptr = orig_load_scnptr = lnnoptr = 0; pagemask = getpagesize () - 1; /* Adjust text/data boundary. */ data_start = (long) start_of_data (); data_start = ADDR_CORRECT (data_start); data_start = data_start & ~pagemask; /* (Down) to page boundary. */ bss_end = ADDR_CORRECT (sbrk (0)) + pagemask; bss_end &= ~ pagemask; /* Adjust data/bss boundary. */ if (bss_start != 0) { bss_start = (ADDR_CORRECT (bss_start) + pagemask); /* (Up) to page bdry. */ bss_start &= ~ pagemask; if (bss_start > bss_end) { ERROR1 ("unexec: Specified bss_start (%u) is past end of program", bss_start); } } else bss_start = bss_end; if (data_start > bss_start) /* Can't have negative data size. */ { ERROR2 ("unexec: data_start (%u) can't be greater than bss_start (%u)", data_start, bss_start); } /* Salvage as much info from the existing file as possible */ f_thdr = NULL; f_dhdr = NULL; f_bhdr = NULL; f_lhdr = NULL; f_tchdr = NULL; f_dbhdr = NULL; f_xhdr = NULL; if (a_out >= 0) { if (read (a_out, &f_hdr, sizeof (f_hdr)) != sizeof (f_hdr)) { PERROR (a_name); } if (f_hdr.f_opthdr > 0) { if (read (a_out, &f_ohdr, sizeof (f_ohdr)) != sizeof (f_ohdr)) { PERROR (a_name); } } if (f_hdr.f_nscns > MAX_SECTIONS) { ERROR0 ("unexec: too many section headers -- increase MAX_SECTIONS"); } /* Loop through section headers */ for (scns = 0; scns < f_hdr.f_nscns; scns++) { struct scnhdr *s = §ion[scns]; if (read (a_out, s, sizeof (*s)) != sizeof (*s)) { PERROR (a_name); } #define CHECK_SCNHDR(ptr, name, flags) \ if (strcmp(s->s_name, name) == 0) { \ if (s->s_flags != flags) { \ fprintf(stderr, "unexec: %lx flags where %x expected in %s section.\n", \ (unsigned long)s->s_flags, flags, name); \ } \ if (ptr) { \ fprintf(stderr, "unexec: duplicate section header for section %s.\n", \ name); \ } \ ptr = s; \ } CHECK_SCNHDR(f_thdr, _TEXT, STYP_TEXT); CHECK_SCNHDR(f_dhdr, _DATA, STYP_DATA); CHECK_SCNHDR(f_bhdr, _BSS, STYP_BSS); CHECK_SCNHDR(f_lhdr, _LOADER, STYP_LOADER); CHECK_SCNHDR(f_dbhdr, _DEBUG, STYP_DEBUG); CHECK_SCNHDR(f_tchdr, _TYPCHK, STYP_TYPCHK); CHECK_SCNHDR(f_xhdr, _EXCEPT, STYP_EXCEPT); } if (f_thdr == 0) { ERROR1 ("unexec: couldn't find \"%s\" section", (int) _TEXT); } if (f_dhdr == 0) { ERROR1 ("unexec: couldn't find \"%s\" section", (int) _DATA); } if (f_bhdr == 0) { ERROR1 ("unexec: couldn't find \"%s\" section", (int) _BSS); } } else { ERROR0 ("can't build a COFF file from scratch yet"); } orig_data_scnptr = f_dhdr->s_scnptr; orig_load_scnptr = f_lhdr ? f_lhdr->s_scnptr : 0; /* Now we alter the contents of all the f_*hdr variables to correspond to what we want to dump. */ /* Indicate that the reloc information is no longer valid for ld (bind); we only update it enough to fake out the exec-time loader. */ f_hdr.f_flags |= (F_RELFLG | F_EXEC); f_ohdr.dsize = bss_start - f_ohdr.data_start; f_ohdr.bsize = bss_end - bss_start; f_dhdr->s_size = f_ohdr.dsize; f_bhdr->s_size = f_ohdr.bsize; f_bhdr->s_paddr = f_ohdr.data_start + f_ohdr.dsize; f_bhdr->s_vaddr = f_ohdr.data_start + f_ohdr.dsize; /* fix scnptr's */ { ulong ptr = section[0].s_scnptr; bias = -1; for (scns = 0; scns < f_hdr.f_nscns; scns++) { struct scnhdr *s = §ion[scns]; if (s->s_flags & STYP_PAD) /* .pad sections omitted in AIX 4.1 */ { /* * the text_start should probably be o_algntext but that doesn't * seem to change */ if (f_ohdr.text_start != 0) /* && scns != 0 */ { s->s_size = 512 - (ptr % 512); if (s->s_size == 512) s->s_size = 0; } s->s_scnptr = ptr; } else if (s->s_flags & STYP_DATA) s->s_scnptr = ptr; else if (!(s->s_flags & (STYP_TEXT | STYP_BSS))) { if (bias == -1) /* if first section after bss */ bias = ptr - s->s_scnptr; s->s_scnptr += bias; ptr = s->s_scnptr; } ptr = ptr + s->s_size; } } /* fix other pointers */ for (scns = 0; scns < f_hdr.f_nscns; scns++) { struct scnhdr *s = §ion[scns]; if (s->s_relptr != 0) { s->s_relptr += bias; } if (s->s_lnnoptr != 0) { if (lnnoptr == 0) lnnoptr = s->s_lnnoptr; s->s_lnnoptr += bias; } } if (f_hdr.f_symptr > 0L) { f_hdr.f_symptr += bias; } text_scnptr = f_thdr->s_scnptr; data_scnptr = f_dhdr->s_scnptr; load_scnptr = f_lhdr ? f_lhdr->s_scnptr : 0; if (write (new_, &f_hdr, sizeof (f_hdr)) != sizeof (f_hdr)) { PERROR (new_name); } if (f_hdr.f_opthdr > 0) { if (write (new_, &f_ohdr, sizeof (f_ohdr)) != sizeof (f_ohdr)) { PERROR (new_name); } } for (scns = 0; scns < f_hdr.f_nscns; scns++) { struct scnhdr *s = §ion[scns]; if (write (new_, s, sizeof (*s)) != sizeof (*s)) { PERROR (new_name); } } return (0); } /* **************************************************************** * * Copy the text and data segments from memory to the new a.out */ static int copy_text_and_data (int new_) { char *end; char *ptr; lseek (new_, (long) text_scnptr, SEEK_SET); ptr = start_of_text () + text_scnptr; end = ptr + f_ohdr.tsize; write_segment (new_, ptr, end); lseek (new_, (long) data_scnptr, SEEK_SET); ptr = (char *) f_ohdr.data_start; end = ptr + f_ohdr.dsize; write_segment (new_, ptr, end); return 0; } #define UnexBlockSz (1<<12) /* read/write block size */ static void write_segment (int new_, char *ptr, char *end) { int i, nwrite, ret; char buf[80]; char zeros[UnexBlockSz]; for (i = 0; ptr < end;) { /* distance to next block. */ nwrite = (((int) ptr + UnexBlockSz) & -UnexBlockSz) - (int) ptr; /* But not beyond specified end. */ if (nwrite > end - ptr) nwrite = end - ptr; ret = write (new_, ptr, nwrite); /* If write gets a page fault, it means we reached a gap between the old text segment and the old data segment. This gap has probably been remapped into part of the text segment. So write zeros for it. */ if (ret == -1 && errno == EFAULT) { memset (zeros, 0, nwrite); write (new_, zeros, nwrite); } else if (nwrite != ret) { sprintf (buf, "unexec write failure: addr 0x%lx, fileno %d, size 0x%x, wrote 0x%x, errno %d", (unsigned long)ptr, new_, nwrite, ret, errno); PERROR (buf); } i += nwrite; ptr += nwrite; } } /* **************************************************************** * copy_sym * * Copy the relocation information and symbol table from the a.out to the new */ static int copy_sym (int new_, int a_out, char *a_name, char *new_name) { char page[UnexBlockSz]; int n; if (a_out < 0) return 0; if (orig_load_scnptr == 0L) return 0; if (lnnoptr && lnnoptr < orig_load_scnptr) /* if there is line number info */ lseek (a_out, lnnoptr, SEEK_SET); /* start copying from there */ else lseek (a_out, orig_load_scnptr, SEEK_SET); /* Position a.out to symtab. */ while ((n = read (a_out, page, sizeof (page))) > 0) { if (write (new_, page, n) != n) { PERROR (new_name); } } if (n < 0) { PERROR (a_name); } return 0; } /* **************************************************************** * mark_x * * After successfully building the new a.out, mark it executable */ static void mark_x (char *name) { struct stat sbuf; int um; int new_ = 0; /* for PERROR */ um = umask (777); umask (um); if (stat (name, &sbuf) == -1) { PERROR (name); } sbuf.st_mode |= 0111 & ~um; if (chmod (name, sbuf.st_mode) == -1) PERROR (name); } static int adjust_lnnoptrs (int UNUSED (writedesc), int UNUSED (readdesc), char *new_name) { int nsyms; int naux; int new_; struct syment symentry; union auxent auxentry; if (!lnnoptr || !f_hdr.f_symptr) return 0; if ((new_ = open (new_name, O_RDWR)) < 0) { PERROR (new_name); return -1; } lseek (new_, f_hdr.f_symptr, SEEK_SET); for (nsyms = 0; nsyms < f_hdr.f_nsyms; nsyms++) { read (new_, &symentry, SYMESZ); if (symentry.n_sclass == C_BINCL || symentry.n_sclass == C_EINCL) { symentry.n_value += bias; lseek (new_, -SYMESZ, SEEK_CUR); write (new_, &symentry, SYMESZ); } for (naux = symentry.n_numaux; naux-- != 0; ) { read (new_, &auxentry, AUXESZ); nsyms++; if (naux != 0 /* skip csect auxentry (last entry) */ && (symentry.n_sclass == C_EXT || symentry.n_sclass == C_HIDEXT)) { auxentry.x_sym.x_fcnary.x_fcn.x_lnnoptr += bias; lseek (new_, -AUXESZ, SEEK_CUR); write (new_, &auxentry, AUXESZ); } } } close (new_); return 0; } static int unrelocate_symbols (int new_, int a_out, char *a_name, char *new_name) { int i; LDHDR ldhdr; LDREL ldrel; ulong t_reloc = (ulong) &_text - f_ohdr.text_start; #ifndef ALIGN_DATA_RELOC ulong d_reloc = (ulong) &_data - f_ohdr.data_start; #else /* This worked (and was needed) before AIX 4.2. I have no idea why. -- Mike */ ulong d_reloc = (ulong) &_data - ALIGN(f_ohdr.data_start, 2); #endif int * p; if (load_scnptr == 0) return 0; lseek (a_out, orig_load_scnptr, SEEK_SET); if (read (a_out, &ldhdr, sizeof (ldhdr)) != sizeof (ldhdr)) { PERROR (new_name); } #define SYMNDX_TEXT 0 #define SYMNDX_DATA 1 #define SYMNDX_BSS 2 for (i = 0; i < ldhdr.l_nreloc; i++) { lseek (a_out, orig_load_scnptr + LDHDRSZ + LDSYMSZ*ldhdr.l_nsyms + LDRELSZ*i, SEEK_SET); if (read (a_out, &ldrel, LDRELSZ) != LDRELSZ) { PERROR (a_name); } /* move the BSS loader symbols to the DATA segment */ if (ldrel.l_symndx == SYMNDX_BSS) { ldrel.l_symndx = SYMNDX_DATA; lseek (new_, load_scnptr + LDHDRSZ + LDSYMSZ*ldhdr.l_nsyms + LDRELSZ*i, SEEK_SET); if (write (new_, &ldrel, LDRELSZ) != LDRELSZ) { PERROR (new_name); } } if (ldrel.l_rsecnm == f_ohdr.o_sndata) { int orig_int; lseek (a_out, orig_data_scnptr + (ldrel.l_vaddr - f_ohdr.data_start), SEEK_SET); if (read (a_out, (void *) &orig_int, sizeof (orig_int)) != sizeof (orig_int)) { PERROR (a_name); } p = (int *) (ldrel.l_vaddr + d_reloc); switch (ldrel.l_symndx) { case SYMNDX_TEXT: orig_int = * p - t_reloc; break; case SYMNDX_DATA: case SYMNDX_BSS: orig_int = * p - d_reloc; break; } if (orig_int != * p) { lseek (new_, data_scnptr + (ldrel.l_vaddr - f_ohdr.data_start), SEEK_SET); if (write (new_, (void *) &orig_int, sizeof (orig_int)) != sizeof (orig_int)) { PERROR (new_name); } } } } return 0; }