From: Mattias Engdegård Date: Sun, 26 May 2024 09:13:50 +0000 (+0200) Subject: Clean up legacy bytecode loading X-Git-Url: http://git.eshelyaron.com/gitweb/?a=commitdiff_plain;h=bf003feccae9f07faef33ccb0a457e1d1a7cf51a;p=emacs.git Clean up legacy bytecode loading Complete some unfinished work from the elimination of lazy bytecode loading in 9bcc9690a8: remove code which became unreachable by that reform. This simplifies some of the lower-level parts of the reader which helps performance somewhat. * src/doc.c (read_bytecode_pointer, read_bytecode_char) (read_doc_string): Remove. (get_doc_string): Remove last argument, all callers adapted. Simplify. * src/lread.c (readbyte_for_lambda, readbyte_from_string) (readbyte_for_lambda, readbyte_from_string): Remove. (readchar, unreadchar): Remove unused code paths for Qlambda and cons arguments. (bytecode_from_rev_list): Remove last use of a quirky code path, decoding force-loaded lazy bytecode explicitly. (cherry picked from commit 236034a24d0e93e1796224e2ff4a14819fcfd559) --- diff --git a/src/doc.c b/src/doc.c index 36633a920c6..f516db3bbcc 100644 --- a/src/doc.c +++ b/src/doc.c @@ -74,24 +74,8 @@ along with GNU Emacs. If not, see . */ static char *get_doc_string_buffer; static ptrdiff_t get_doc_string_buffer_size; -static unsigned char *read_bytecode_pointer; - static char const sibling_etc[] = "../etc/"; -/* `readchar' in lread.c calls back here to fetch the next byte. - If UNREADFLAG is 1, we unread a byte. */ - -int -read_bytecode_char (bool unreadflag) -{ - if (unreadflag) - { - read_bytecode_pointer--; - return 0; - } - return *read_bytecode_pointer++; -} - #ifdef USE_ANDROID_ASSETS /* Like `close_file_unwind'. However, PTR is a pointer to an Android @@ -120,15 +104,10 @@ close_file_unwind_android_fd (void *ptr) (e.g. because the file has been modified and the location is stale), return nil. - If UNIBYTE, always make a unibyte string. - - If DEFINITION, assume this is for reading - a dynamic function definition; convert the bytestring - and the constants vector with appropriate byte handling, - and return a cons cell. */ + If UNIBYTE, always make a unibyte string. */ Lisp_Object -get_doc_string (Lisp_Object filepos, bool unibyte, bool definition) +get_doc_string (Lisp_Object filepos, bool unibyte) { char *from, *to, *name, *p, *p1; Lisp_Object file, pos; @@ -312,14 +291,6 @@ Invalid data in documentation file -- %c followed by code %03o", *to++ = *from++; } - /* If DEFINITION, read from this buffer - the same way we would read bytes from a file. */ - if (definition) - { - read_bytecode_pointer = (unsigned char *) get_doc_string_buffer + offset; - return Fread (Qlambda); - } - if (unibyte) return make_unibyte_string (get_doc_string_buffer + offset, to - (get_doc_string_buffer + offset)); @@ -336,16 +307,6 @@ Invalid data in documentation file -- %c followed by code %03o", } } -/* Get a string from position FILEPOS and pass it through the Lisp reader. - We use this for fetching the bytecode string and constants vector - of a compiled function from the .elc file. */ - -Lisp_Object -read_doc_string (Lisp_Object filepos) -{ - return get_doc_string (filepos, 0, 1); -} - static bool reread_doc_file (Lisp_Object file) { @@ -406,7 +367,7 @@ string is passed through `substitute-command-keys'. */) if (FIXNUMP (doc) || CONSP (doc)) { Lisp_Object tem; - tem = get_doc_string (doc, 0, 0); + tem = get_doc_string (doc, 0); if (NILP (tem) && try_reload) { /* The file is newer, we need to reset the pointers. */ @@ -481,7 +442,7 @@ aren't strings. */) if (FIXNUMP (tem) || (CONSP (tem) && FIXNUMP (XCDR (tem)))) { Lisp_Object doc = tem; - tem = get_doc_string (tem, 0, 0); + tem = get_doc_string (tem, 0); if (NILP (tem) && try_reload) { /* The file is newer, we need to reset the pointers. */ diff --git a/src/lisp.h b/src/lisp.h index 534a36499f1..f8d59b1e9fd 100644 --- a/src/lisp.h +++ b/src/lisp.h @@ -5273,10 +5273,8 @@ extern void set_initial_environment (void); extern void syms_of_callproc (void); /* Defined in doc.c. */ -extern Lisp_Object read_doc_string (Lisp_Object); -extern Lisp_Object get_doc_string (Lisp_Object, bool, bool); +extern Lisp_Object get_doc_string (Lisp_Object, bool); extern void syms_of_doc (void); -extern int read_bytecode_char (bool); /* Defined in bytecode.c. */ extern void syms_of_bytecode (void); diff --git a/src/lread.c b/src/lread.c index c92b2ede932..233f08b0727 100644 --- a/src/lread.c +++ b/src/lread.c @@ -256,15 +256,13 @@ static Lisp_Object oblookup_considering_shorthand (Lisp_Object, const char *, ptrdiff_t *); -/* Functions that read one byte from the current source READCHARFUN +/* Function that reads one byte from the current source READCHARFUN or unreads one byte. If the integer argument C is -1, it returns one read byte, or -1 when there's no more byte in the source. If C is 0 or positive, it unreads C, and the return value is not interesting. */ -static int readbyte_for_lambda (int, Lisp_Object); static int readbyte_from_file (int, Lisp_Object); -static int readbyte_from_string (int, Lisp_Object); /* Handle unreading and rereading of characters. Write READCHAR to read a character, @@ -278,8 +276,8 @@ static int readbyte_from_string (int, Lisp_Object); /* Same as READCHAR but set *MULTIBYTE to the multibyteness of the source. */ #define READCHAR_REPORT_MULTIBYTE(multibyte) readchar (readcharfun, multibyte) -/* When READCHARFUN is Qget_file_char, Qget_emacs_mule_file_char, - Qlambda, or a cons, we use this to keep an unread character because +/* When READCHARFUN is Qget_file_char or Qget_emacs_mule_file_char, + we use this to keep an unread character because a file stream can't handle multibyte-char unreading. The value -1 means that there's no unread character. */ static int unread_char = -1; @@ -365,12 +363,6 @@ readchar (Lisp_Object readcharfun, bool *multibyte) return c; } - if (EQ (readcharfun, Qlambda)) - { - readbyte = readbyte_for_lambda; - goto read_multibyte; - } - if (EQ (readcharfun, Qget_file_char)) { eassert (infile); @@ -400,20 +392,6 @@ readchar (Lisp_Object readcharfun, bool *multibyte) return c; } - if (CONSP (readcharfun) && STRINGP (XCAR (readcharfun))) - { - /* This is the case that read_vector is reading from a unibyte - string that contains a byte sequence previously skipped - because of #@NUMBER. The car part of readcharfun is that - string, and the cdr part is a value of readcharfun given to - read_vector. */ - readbyte = readbyte_from_string; - eassert (infile); - if (EQ (XCDR (readcharfun), Qget_emacs_mule_file_char)) - emacs_mule_encoding = 1; - goto read_multibyte; - } - if (EQ (readcharfun, Qget_emacs_mule_file_char)) { readbyte = readbyte_from_file; @@ -545,14 +523,6 @@ unreadchar (Lisp_Object readcharfun, int c) read_from_string_index_byte = string_char_to_byte (readcharfun, read_from_string_index); } - else if (CONSP (readcharfun) && STRINGP (XCAR (readcharfun))) - { - unread_char = c; - } - else if (EQ (readcharfun, Qlambda)) - { - unread_char = c; - } else if (FROM_FILE_P (readcharfun)) { unread_char = c; @@ -561,13 +531,6 @@ unreadchar (Lisp_Object readcharfun, int c) call1 (readcharfun, make_fixnum (c)); } -static int -readbyte_for_lambda (int c, Lisp_Object readcharfun) -{ - return read_bytecode_char (c >= 0); -} - - static int readbyte_from_stdio (void) { @@ -638,26 +601,6 @@ readbyte_from_file (int c, Lisp_Object readcharfun) return readbyte_from_stdio (); } -static int -readbyte_from_string (int c, Lisp_Object readcharfun) -{ - Lisp_Object string = XCAR (readcharfun); - - if (c >= 0) - { - read_from_string_index--; - read_from_string_index_byte - = string_char_to_byte (string, read_from_string_index); - } - - return (read_from_string_index < read_from_string_limit - ? fetch_string_char_advance (string, - &read_from_string_index, - &read_from_string_index_byte) - : -1); -} - - /* Signal Qinvalid_read_syntax error. S is error string of length N (if > 0) */ @@ -3517,12 +3460,18 @@ bytecode_from_rev_list (Lisp_Object elems, Lisp_Object readcharfun) /* Lazily-loaded bytecode is represented by the constant slot being nil and the bytecode slot a (lazily loaded) string containing the - print representation of (BYTECODE . CONSTANTS). Unpack the - pieces by coerceing the string to unibyte and reading the result. */ + print representation of (BYTECODE . CONSTANTS). */ if (NILP (vec[CLOSURE_CONSTANTS]) && STRINGP (vec[CLOSURE_CODE])) { Lisp_Object enc = vec[CLOSURE_CODE]; - Lisp_Object pair = Fread (Fcons (enc, readcharfun)); + eassert (!STRING_MULTIBYTE (enc)); + /* The string (always unibyte) must be decoded to be parsed. */ + enc = Fdecode_coding_string (enc, + EQ (readcharfun, + Qget_emacs_mule_file_char) + ? Qemacs_mule : Qutf_8_emacs, + Qt, Qnil); + Lisp_Object pair = Fread (enc); if (!CONSP (pair)) invalid_syntax ("Invalid byte-code object", readcharfun); @@ -3772,7 +3721,7 @@ get_lazy_string (Lisp_Object val) && !(pos >= ss->position && pos < ss->position + ss->length)) ss++; if (ss >= ssend) - return get_doc_string (val, 1, 0); + return get_doc_string (val, 1); ptrdiff_t start = pos - ss->position; char *str = ss->string;