changelo

GNU 系统开发优化 C 语言程序的应用程序

Wed Jun 16 15:32:36 1993  Joseph Arceneaux  ([e-mail])

	* Version 1.8 released.

	* makefile.in: New variable VMSFILES describing the files needed
	for indent under VMS.
        Removed comments.texinfo from list of files to tar.
	* configure-6-93:
	* configure.in-693: Copies of the configure script and its input
	file from the autoconf directory.  The reliability of the ones in
	the autoconf directory is questionable.  `configure' and
	`configure.in' are now links to these files.

	* version.h: Comments rewritten.

Tue Jun 15 10:14:47 1993  Joseph Arceneaux  ([e-mail])

	* Thanks to [e-mail] for the VMS patches.
	* io.c (read_file): Use SYS_READ, conditional expansion, instead
	of read.
	(vms_read): Substitute for `read' under VMS.

	* backup.c: Set `simple_backup_suffix' from new define
	BACKUP_SUFFIX_STR.  Define this (to "~") if undefined.
	Likewise, BACKUP_SUFFIX_CHAR and BACKUP_SUFFIX_FORMAT.
	* sys.h: Conditional defines for VMS added.
	* args.c (set_profile): Use new define PROFILE_FORMAT for
	formatting homedir/profile path.

Mon Jun 14 11:37:14 1993  Joseph Arceneaux  ([e-mail])

	* indent.texinfo (Comments): Section completely rewritten.  Old
	 text in file old-comments.texinfo.

	* indent.texinfo (Statements): Explained how -bl and -br affect
	struct declarations.

	* indent.c (indent): In case newline, modified conditional
	determining whether or not to call `dump_line' to avoid doing so
	after a right brace if user has specified `btype_2'.  This makes
	enums and structs more consistent -- in default mode, the name of
	the struct goes on a separate line from the rbrace and on the same
	line in K&R mode, and does so consistently whether in or outside a
	procedure block.

	* sys.h: Added some defines for compilation under MS-DOS, under
	#ifdef __MSDOS__.  Thanks to [e-mail].

	* io.c (read_file):  Adjust `fileptr.size' to be what `read'
	returned, if different from the results of `stat'.  This is
	because the DOS `read' converts the CR-LF on disc to '\n' in
	memory.

Fri Jun 11 12:39:38 1993  Joseph Arceneaux  ([e-mail])

	* lexi.c (lexi): Accept LL and ULL as suffixes to numbers as well.
	Suggested by [e-mail].

	* indent.c (indent): In case `rbrace', when conditionally setting
	`force_nl', extended the first clause of the if conditional to
	make `rw_decl' equivalent to `rw_struct_like', and also depend on
	`btype_2'.  This keeps the structure name on the same line when
	running with -kr option.
	* parse.c (parse): In case `lbrace', in conditional for adding
	`brace_indent', also consider `btype_2'.

Thu Jun 10 17:21:54 1993  Joseph Arceneaux  ([e-mail])

	* io.c (count_columns): Rewritten and renamed from `count_spaces'.

Wed Jun  9 12:45:07 1993  Joseph Arceneaux  ([e-mail])

	* indent.texinfo (Disabling Formatting): New section describing
	the new feature implemented below in io.c.

	* io.c (compute_code_target): Rewritten to be clearer.  Note that
	the use of `max_col' is inconsistent with it's use elsewhere,
	where it is primarily for use with comment formatting.
	* (fill_buffer): Rewritten completely.  Now handles command
	comments /* *INDENT-OFF* */ and /* *INDENT-ON* */.
	`inhibit_formatting' is now a local variable;  the reference to it
	in `dump_line' has been removed.  Control of formatting is handled
	entirely within `fill_buffer'.

Tue Jun  8 19:03:22 1993  Joseph Arceneaux  ([e-mail])

	* indent.c (indent): In case `lbrace', when resetting
	..->decl_on_line, also reset ->in_decl.  Not doing this had the
	effect that the first statement after the opening '{' of a
	procedure would not be properly continued if it was broken across
	two lines.  There is probably a better place to do this.

	* io.c (dump_line): In the loop which outputs `s_code', removed
	the stupid tab calculations, as well as the undocumented feature
	for printing out the character '0200'.

	* parse.c (parse): In case `lbrace', if last reserved word was
	`rw_struct_like', don't indent for struct if the last token was an
	`rparen'.  This fixes a bug with a struct pointer parameter.

	* regression/TEST: Added new test struct_param.c.

Mon Jun  7 17:01:23 1993  Joseph Arceneaux  ([e-mail])

	* indent.c (indent): When handling an embedded comment, if there
	is code on the line, set `embedded_comment_on_line' to 2,
	otherwise to 1, to communicate to dump_line () if the line starts
	with code or comment text.
	* io.c (dump_line): In section for code (s_code != e_code), if
	`embedded_comment_on_line' is 1, then use comment indentation,
	otherwise, use code indentation.

Fri Jun  4 18:10:54 1993  Joseph Arceneaux  ([e-mail])

	* pr_comment.c (print_comment): When looking at a leading '*' to
	decide if it begins the second line of a starred boxed comment (by
	looking at what column it was in), compare that column to
	`found_column' rather than `start_column'.

	* indent.c (indent): In case preesc, when handling stuff following
	the '#', set `quote' when we encounter the quoting character the
	first time.  Also merged cases '\'' and '"'.

	* regression directory:  Merged all the new comment code tests
	into the standard regression tests.

Wed May 19 12:11:10 1993  Joseph Arceneaux  ([e-mail])

	* pr_comment.c (print_comment): At `end_line:', don't compress
	into newlines if it's the first line (because it has the starting
	delimiter).

	* pr_comment.c (print_comment): Moved checks for `buf_ptr' past
	end outside of while loops (only '\n' and '\0' end the buffer, and
	changed "buf_ptr >= buf_end" to "buf_ptr == buf_end".

	* parse.c: `parser_state_tos' elements `comment_delta' and
	`n_comment_delta' eliminated.
	* indent.h: Ditto.
	* pr_comment.c: Ditto.

	* io.c: Old comment code removed.

	* pr_comment.c (print_comment): New variable `visible_preamble'
	set if the preamble is not whitespace.  Used to make sure lines
	which only have whitespace on them become single newlines.
	Don't use `postfix_blankline_requested' to add newlines.  This
	screwed up the algorithm for dealing with paragraph breaks.
	In the `end_line:' section, only advance `buf_ptr' if it wasn't a
	paragraph break, because in that case we have scanned ahead a
	character to detect "\n\n".

Tue May 11 17:26:18 1993  Joseph Arceneaux  ([e-mail])

	* pr_comment.c (print_comment): For comment formatting, changed
	comparison of `column' to `right_margin' from >= to >.

Tue Apr  6 17:51:41 1993  Joseph Arceneaux  ([e-mail])

	* pr_comment.c (print_comment): Also consider the characters '='
	and '_' to make up the top line of a boxed comment.
	* comments.texinfo: Changed to reflect this.

Wed Feb 24 15:22:51 1993  Joseph Arceneaux  ([e-mail])

	* makefile.in: Added "-O" to CFLAGS.

Thu Feb 11 15:54:10 1993  Joseph Arceneaux  ([e-mail])

	* Version 1.7 released.

	* makefile.in: Renamed from zmakefile.in.
	All "-f zmakefile" arguments to `MAKE' removed.  Dependencies on
	"zmakefile.in" changed to "makefile.in".

	* Makefile: Changed references to "zmakefile" to "makefile".

Wed Feb 10 12:25:04 1993  Joseph Arceneaux  ([e-mail])

	* indent.c (indent): In case preesc, when going through the line,
	consider EOL as terminating a C++ comment.

Mon Feb  8 04:45:06 1993  Joseph Arceneaux  ([e-mail])

	* memcpy.c: Function `memcpy' surrounded by #ifdef USG.  USG is
	only defined if `configure' did not find `bcopy'.  memcpy.o should
	only be loaded if `configure' did not find it.
	* sys.h: If USG not defined, define memcpy to be bcopy.

	* makefile.in: Renamed from zmakefile.in.
	* configure: Use "makefile" rather than zmakefile.  Explicitly
	name `indent' in rule for same.

	* indent.texinfo: Updated copyright and version information.
	Added new section Bugs.  Added "nlps" option.

	* io.c (pad_output): Rewritten.  If `tabsize' is less than one,
	use spaces rather than tabs.

	* sys.h: #define TAB and EOL.
	Changed type of element `size' in struct file_buffer to unsigned
	long.

	* indent.h, io.c: Type of `in_prog_size' changed to unsigned long.

	* indent.c, parse.c, pr_comment.c, lexi.c, io.c: Use new macros
	TAB and EOL instead of '\t' and '\n'.

	* args.c, indent.c, backup.c, io.c, parse.c: Cast integer
	arguments to printf.

	* indent.c:
	* lexi.c:
	* parse.c: Added missing parameters to calls to `diag'.

	* args.c: Added new long option "remove-preprocessor-space".

Fri Feb  5 17:40:08 1993  Joseph Arceneaux  ([e-mail])

	* indent.c (indent): In case semicolon, case lbrace, and case
	rbrace, disabled code which checked paren levels (with
	parser_state_tos->p_l_follow) to enable passing weird stuff to
	macros.  See comments ending "-jla".
	(indent): In case lparen, if in a declaration, don't indent out to
	`dec_ind' if the paren was '['.
	* Removed several global variables which were shadowed by locals
	to function `indent'.

	* args.c: Turned off, by default, "lps" option.  Added "nlps"
	option.

	* io.c (current_column): New function.

	* lexi.c: In case ':', set parser_state_tos->want_blank under
	certain conditions.

Tue Feb  2 14:03:39 1993  Joseph Arceneaux  ([e-mail])

	* indent.c (indent): Handle C++ comments, with special attention
	to the code for case preesc.

Thu Jan 28 17:31:57 1993  Joseph Arceneaux  ([e-mail])

	* parse.c (reset_parser): Initialize parser_state_tos->com_col to
	0.

Wed Jan 27 13:45:16 1993  Joseph Arceneaux  ([e-mail])

	* indent.c (indent): Handle new code type `cplus_comment'.

	* lexi.c (lexi): Handle C++ comments ("//").

	* indent.h: Macros CHECK_CODE_SIZE and CHECK_LAB_SIZE moved to
	indent.c.  Macro CHECK_COM_SIZE moved to pr_comment.c.
	New code, `cplus_comment'.

	* pr_comment.c (pr_comment): Removed variable `just_saw_decl' and
	it's use.

Sat Jan 23 19:36:31 1993  Joseph Arceneaux  ([e-mail])

	* indent.c (indent): Only reset `com_ind' if it's less than or
	equal to 0.  Don't call parse (semicolon) initially.  Skip any
	leading newlines, resetting col to 1. Don't set
	`parser_state_tos->ind_level' and `parser_state_tos->i_l_follow'
	if col > ind_size.

	* parse.c (reset_parser): If tabsize is 0, set it to 1.

Fri Oct  9 21:01:49 1992  Joseph Arceneaux  ([e-mail])

	* indent.c (indent): Restuctured the conditional for setting
	`force_nl' in case rbrace to be clearer.  Also, only check whether
	to set `postfix_blankline_requested' we're not setting `force_nl'.

Wed Oct  7 16:15:25 1992  Joseph Arceneaux  ([e-mail])

	* io.c (diag): Only one format of error message now.  Output it on
	the standard error in any case.

	* indent.c (indent): In case decl, don't call "parse (decl)" if we
	are in a "sizeof".

	* args.c: Make default value for `format_col1_comments' be false,
	to correspond with the "-gnu" spec.

Fri Aug 28 19:15:50 1992  Joseph Arceneaux  ([e-mail])

	* args.c (set_option): Added missing argument to call to
	`set_option'.

Mon Aug 24 12:47:32 1992  Joseph Arceneaux  ([e-mail])

	* sys.h (INLINE): Define this to "__inline__" if __GNUC__ is
	defined, not "inline".  Good thing we have standards...

	* indent.c (indent): New variable `embedded_comment_on_line' set
	if an embedded comment is found.
	* io.c (dump_line): Increment `com_lines' in the case where the
	comment buffer is empty, but `embedded_comment_on_line' is true.
	Clear that variable in both cases.  This makes indent include
	embedded comments in its count of total comment lines.

	* io.c (read_stdin): If the input buffer is realloc'd, update the
	character pointer `p'.

Wed Aug 19 18:25:48 1992  Joseph Arceneaux  ([e-mail])

	* Version 1.6 released.

	* indent.c (indent): In case comma, use `tabsize' rather than 8 to
	determine if we're past max_col.
	In case newline, also dump the line if we are past max_col.

Fri Aug 14 19:08:04 1992  Joseph Arceneaux  ([e-mail])

	* io.c (dump_line): Test `swallow_optional_blanklines' separately
	from (prefix_blankline_requested && not_first_line).  Makes "-sob"
	work properly.

	* indent.c (indent): Intialize several local variables to 0.

Tue Aug 11 18:02:38 1992  Joseph Arceneaux  ([e-mail])

	* indent.c (indent): In case rbrace, add clause to conditional for
	setting `force_nl'.  The clause forces a newline after '}' in most
	cases.
	In case swstmt, set `parser_state_tos->in_decl' to false, since we
	know we're not in a declaration if have just seen "switch" (of
	course, who knows what indent might think about things).  This
	corrects the indentation of a case label following the lbrace of
	the switch.

Mon Aug 10 16:23:54 1992  Joseph Arceneaux  ([e-mail])

	* indent.gperf: "return" generates `rw_return', not `rw_break'.
	How did this happen?  I didn't make this change, and it was not
	present in indent 1.4.  Doubtless this insidious program modifies
	itself destructively...
	* lexi.c (hash, is_reserved): regenerated from indent.gperf.

	* args.c: Make the -gnu setting specify "-bli2" rather than
	"-bli4".

	* lexi.c (lexi): Accept 'F' suffixes on numeric constants.

	* sys.h Require "__GNUC__" to define INLINE, not "__STDC__".

Mon Aug  4 21:22:54 1992  Joseph Arceneaux  ([e-mail])

	* Version 1.5 released.

Mon Aug  3 21:14:40 1992  Joseph Arceneaux  ([e-mail])

	* indent.c (indent): In case rparen, removed the change of Jul 28.
	This is now done if the case mask fails and
	parser_state_tos->in_decl is true and ..->block_init is false.
	The old change broke the "-ncs" feature.  Perhaps this could have
	been done otherwise by using ..->last_u_d.
	Also, decrement `parser_state_tos->paren_depth' immediatly.

	* All source files formatted with ./indent.

	* pr_comment.c (pr_comment): If `parser_state_tos->box_com' is set
	(which means that the comment is presumed to be hand-formatted),
	set `parser_state_tos->n_comment_delta' to
	(1 - parser_state_tos->com_col), rather than using `count_spaces'
	to calculate the space up to the comment column.  Not only is this
	simpler, but the old code used `cur_line' to count those spaces,
	which was pointing to the *NEXT* line in some cases (because some
	code in indent () which ignored newlines (like in the case of
	multi-line comments after #define statments) called fill_buffer ()).
	Another example of just how brain damaged this code is.  For more
	extreme self-abuse, look a the interaction between pr_comment ()
	and dump_line ().

Sun Aug  2 02:10:32 1992  Joseph Arceneaux  ([e-mail])

	* pr_comment.c (pr_comment): When computing
	`parser_state_tos->com_col' for beyond the normal comment
	position, use (tabsize - 1) instead of 7.

Wed Jul 29 22:50:45 1992  Joseph Arceneaux  ([e-mail])

	* lexi.c (lexi): If `buf_ptr' == '(' after scanning an alphanum
	which is not a keyword, add clause that paren_depth must be 0
	before considering this as a possible function definition.

Tue Jul 28 17:17:08 1992  Joseph Arceneaux  ([e-mail])

	* parse.c (parse): In case lbrace, handle state of being in a decl