1ChangeLog entry:
2
3Thanks to Paul Eggert who suggested using better random numbers as
4well as using the base62 format for compactness and provided the
5sample divide_by and convert functions used here.
6
72005-09-29  Mark D. Baushke  <mdb@gnu.org>
8
9	* man/rcsfile.5in: Document new commitid delta phrase.
10	* man/rcsfile.5: Regenerated.
11
12	* src/ci.c (RANDOM_BYTES, COMMITID_RAW_SIZE): New constants.
13	(mainProg): Add commitid to delta records. Use
14	random data and represent in base62 or fall back to using the
15	same basic format construction as is used by CVS and CVSNT.
16	(divide_by): New function used by convert.
17	(convert): New fucntion to convert to base62.
18	* rcsbase.h (commitidsize): Room for base62 encoded block or
19	32bit pid plus a 32bit time rendered as hex plus one
20	NUL byte round up to 64.
21	(struct hshentry): Add new commitid field.
22	* src/rcsgen.c (putdelta): Preserve old commitid entries.
23	* src/rcssyn.c (Kcommitid): New global constant keyword.
24	(getdelta): Add optional parsing for it.
25	* src/rlog.c (putadelta): Print it out.
26
27Index:man/rcsfile.5
28--- man/rcsfile.5~	1995-06-16 06:58:26.000000000 +0000
29+++ man/rcsfile.5	2005-09-27 20:53:01.023504000 +0000
30@@ -1,4 +1,4 @@
31-.lf 1 ./rcsfile.5in
32+.lf 1 rcsfile.5in
33 .\" Set p to 1 if your formatter can handle pic output.
34 .if t .nr p 1
35 .de Id
36@@ -69,6 +69,7 @@ nonterminal symbols are in
37 		\f3state\fP	{\f2id\fP}\f3;\fP
38 		\f3branches\fP	{\f2num\fP}*\f3;\fP
39 		\f3next\fP	{\f2num\fP}\f3;\fP
40+		{ \f3commitid\fP \f2id\fP\f3;\fP }
41 		{ \f2newphrase\fP }*
42 .LP
43 \f2desc\fP	::=	\f3desc\fP	\f2string\fP
44@@ -128,6 +129,18 @@ and all the digits of years thereafter.
45 Dates use the Gregorian calendar; times use UTC.
46 .PP
47 The
48+.I commitid
49+is followed by an
50+.I id
51+token. This token is intended to be unique across
52+multiple files and is used to help group files as
53+being a part of the same logical commit.
54+This token must uniquely identify the commit
55+operation that was applied to a set of RCS files.
56+In particular, it must be unique among all the
57+commitids in this file.
58+.PP
59+The
60 .I newphrase
61 productions in the grammar are reserved for future extensions
62 to the format of \*r files.
63@@ -230,7 +243,7 @@ The following diagram shows an example o
64 .fi
65 .\}
66 .if \np \{\
67-.lf 232
68+.lf 245
69 .PS 4.250i 3.812i
70 .\" -2.0625 -4.25 1.75 0
71 .\" 0.000i 4.250i 3.812i 0.000i
72@@ -239,7 +252,7 @@ The following diagram shows an example o
73 .nr 0x 1
74 \h'3.812i'
75 .sp -1
76-.lf 242
77+.lf 255
78 \h'2.062i-(\w'Head'u/2u)'\v'0.125i-(0v/2u)+0v+0.22m'Head
79 .sp -1
80 \h'2.062i'\v'0.250i'\D'l0.000i 0.500i'
81@@ -256,7 +269,7 @@ The following diagram shows an example o
82 .sp -1
83 \h'1.688i'\v'0.750i'\D'l0.000i 0.500i'
84 .sp -1
85-.lf 244
86+.lf 257
87 \h'2.062i-(\w'2.1'u/2u)'\v'1.000i-(0v/2u)+0v+0.22m'2.1
88 .sp -1
89 \h'2.062i'\v'1.250i'\D'l0.000i 0.500i'
90@@ -265,7 +278,7 @@ The following diagram shows an example o
91 .sp -1
92 \h'2.062i'\v'1.750i'\D'l-0.025i -0.100i'
93 .sp -1
94-.lf 246
95+.lf 259
96 \h'2.062i-(\w'1.3'u/2u)'\v'2.000i-(1v/2u)+0v+0.22m'1.3
97 .sp -1
98 \h'2.062i'\v'2.250i'\D'l-0.375i -0.500i'
99@@ -280,7 +293,7 @@ The following diagram shows an example o
100 .sp -1
101 \h'1.375i'\v'1.500i'\D'l0.025i 0.100i'
102 .sp -1
103-.lf 249
104+.lf 262
105 \h'1.375i-(\w'1.3.1.1'u/2u)'\v'1.250i-(1v/2u)+1v+0.22m'1.3.1.1
106 .sp -1
107 \h'1.375i'\v'1.000i'\D'l-0.375i 0.500i'
108@@ -295,7 +308,7 @@ The following diagram shows an example o
109 .sp -1
110 \h'2.062i'\v'2.750i'\D'l-0.025i -0.100i'
111 .sp -1
112-.lf 252
113+.lf 265
114 \h'2.062i-(\w'1.2'u/2u)'\v'3.000i-(1v/2u)+0v+0.22m'1.2
115 .sp -1
116 \h'2.062i'\v'3.250i'\D'l-0.375i -0.500i'
117@@ -310,7 +323,7 @@ The following diagram shows an example o
118 .sp -1
119 \h'0.375i'\v'2.500i'\D'l0.025i 0.100i'
120 .sp -1
121-.lf 255
122+.lf 268
123 \h'0.375i-(\w'1.2.1.1'u/2u)'\v'2.250i-(1v/2u)+1v+0.22m'1.2.1.1
124 .sp -1
125 \h'0.375i'\v'2.000i'\D'l-0.375i 0.500i'
126@@ -325,7 +338,7 @@ The following diagram shows an example o
127 .sp -1
128 \h'0.375i'\v'1.500i'\D'l0.025i 0.100i'
129 .sp -1
130-.lf 257
131+.lf 270
132 \h'0.375i-(\w'1.2.1.3'u/2u)'\v'1.250i-(1v/2u)+1v+0.22m'1.2.1.3
133 .sp -1
134 \h'0.375i'\v'1.000i'\D'l-0.375i 0.500i'
135@@ -340,7 +353,7 @@ The following diagram shows an example o
136 .sp -1
137 \h'2.750i'\v'2.500i'\D'l0.025i 0.100i'
138 .sp -1
139-.lf 261
140+.lf 274
141 \h'2.750i-(\w'1.2.2.1'u/2u)'\v'2.250i-(1v/2u)+1v+0.22m'1.2.2.1
142 .sp -1
143 \h'2.750i'\v'2.000i'\D'l-0.375i 0.500i'
144@@ -355,7 +368,7 @@ The following diagram shows an example o
145 .sp -1
146 \h'3.438i'\v'1.250i'\D'l0.025i 0.100i'
147 .sp -1
148-.lf 264
149+.lf 277
150 \h'3.438i-(\w'\s-21.2.2.1.1.1\s0'u/2u)'\v'1.000i-(1v/2u)+1v+0.22m'\s-21.2.2.1.1.1\s0
151 .sp -1
152 \h'3.438i'\v'0.750i'\D'l-0.375i 0.500i'
153@@ -370,7 +383,7 @@ The following diagram shows an example o
154 .sp -1
155 \h'2.750i'\v'1.500i'\D'l0.025i 0.100i'
156 .sp -1
157-.lf 267
158+.lf 280
159 \h'2.750i-(\w'1.2.2.2'u/2u)'\v'1.250i-(1v/2u)+1v+0.22m'1.2.2.2
160 .sp -1
161 \h'2.750i'\v'1.000i'\D'l-0.375i 0.500i'
162@@ -385,7 +398,7 @@ The following diagram shows an example o
163 .sp -1
164 \h'2.062i'\v'3.750i'\D'l-0.025i -0.100i'
165 .sp -1
166-.lf 270
167+.lf 283
168 \h'2.062i-(\w'1.1'u/2u)'\v'4.000i-(1v/2u)+0v+0.22m'1.1
169 .sp -1
170 \h'2.062i'\v'4.250i'\D'l-0.375i -0.500i'
171@@ -398,9 +411,9 @@ The following diagram shows an example o
172 .if \n(00 .fi
173 .br
174 .nr 0x 0
175-.lf 271
176+.lf 284
177 .PE
178-.lf 272
179+.lf 285
180 .\}
181 .PP
182 .SH IDENTIFICATION
183Index:man/rcsfile.5in
184--- man/rcsfile.5in~	1995-06-05 08:28:35.000000000 +0000
185+++ man/rcsfile.5in	2005-09-27 20:52:46.424504000 +0000
186@@ -68,6 +68,7 @@ nonterminal symbols are in
187 		\f3state\fP	{\f2id\fP}\f3;\fP
188 		\f3branches\fP	{\f2num\fP}*\f3;\fP
189 		\f3next\fP	{\f2num\fP}\f3;\fP
190+		{ \f3commitid\fP \f2id\fP\f3;\fP }
191 		{ \f2newphrase\fP }*
192 .LP
193 \f2desc\fP	::=	\f3desc\fP	\f2string\fP
194@@ -127,6 +128,18 @@ and all the digits of years thereafter.
195 Dates use the Gregorian calendar; times use UTC.
196 .PP
197 The
198+.I commitid
199+is followed by an
200+.I id
201+token. This token is intended to be unique across
202+multiple files and is used to help group files as
203+being a part of the same logical commit.
204+This token must uniquely identify the commit
205+operation that was applied to a set of RCS files.
206+In particular, it must be unique among all the
207+commitids in this file.
208+.PP
209+The
210 .I newphrase
211 productions in the grammar are reserved for future extensions
212 to the format of \*r files.
213Index:src/rcsbase.h
214--- src/rcsbase.h~	1995-06-16 06:19:24.000000000 +0000
215+++ src/rcsbase.h	2005-09-28 21:47:51.490505000 +0000
216@@ -222,6 +222,11 @@ Report problems and direct all questions
217                               /* 1 sets the default locking to strict;      */
218                               /* used in production environments.           */
219
220+/* base64_encode(128 random bits) needs 24 bytes + 1 for NUL */
221+/* time_t may be 64bits on some machines needs 16 bytes + 1 as hex */
222+#define commitidsize	   64 /* time+1+base64(128bits)+1 | pid+time+rand+1 */
223+#define urandom_dev "/dev/urandom"
224+
225 #define yearlength	   16 /* (good through AD 9,999,999,999,999,999)    */
226 #define datesize (yearlength+16)	/* size of output of time2date */
227 #define RCSTMPPREFIX '_' /* prefix for temp files in working dir  */
228@@ -358,6 +363,7 @@ struct hshentry {
229 	char const	  * lockedby; /* who locks the revision		    */
230 	char const	  * state;    /* state of revision (Exp by default) */
231 	char const	  * name;     /* name (if any) by which retrieved   */
232+	char const        * commitid; /* text string to associate commits   */
233 	struct cbuf	    log;      /* log message requested at checkin   */
234         struct branchhead * branches; /* list of first revisions on branches*/
235 	struct cbuf	    ig;	      /* ignored phrases in admin part	    */
236@@ -662,6 +668,7 @@ extern int               TotalDeltas;
237 extern char const *const expand_names[];
238 extern char const
239 	Kaccess[], Kauthor[], Kbranch[], Kcomment[],
240+	Kcommitid[],
241 	Kdate[], Kdesc[], Kexpand[], Khead[], Klocks[], Klog[],
242 	Knext[], Kstate[], Kstrict[], Ksymbols[], Ktext[];
243 void unexpected_EOF P((void)) exiting;
244Index:src/ci.c
245--- src/ci.c~	1995-06-16 06:19:24.000000000 +0000
246+++ src/ci.c	2005-09-29 21:57:57.814504000 +0000
247@@ -262,6 +262,10 @@ static void cleanup P((void));
248 static void incnum P((char const*,struct buf*));
249 static void addassoclst P((int,char const*));
250
251+enum {RANDOM_BYTES = 8};
252+enum {COMMITID_RAW_SIZE = (sizeof(time_t) + RANDOM_BYTES)};
253+static void convert P((char const input[COMMITID_RAW_SIZE], char *output));
254+
255 static FILE *exfile;
256 static RILE *workptr;			/* working file pointer		*/
257 static struct buf newdelnum;		/* new revision number		*/
258@@ -285,6 +289,7 @@ mainProg(ciId, "ci", "$Id: ci.c,v 5.30 1
259 	char olddate[datesize];
260 	char newdatebuf[datesize + zonelenmax];
261 	char targetdatebuf[datesize + zonelenmax];
262+	char commitid[commitidsize];
263 	char *a, **newargv, *textfile;
264 	char const *author, *krev, *rev, *state;
265 	char const *diffname, *expname;
266@@ -309,6 +314,45 @@ mainProg(ciId, "ci", "$Id: ci.c,v 5.30 1
267 	suffixes = X_DEFAULT;
268 	nextassoc = &assoclst;
269
270+	{
271+		char buf[COMMITID_RAW_SIZE] = { 0, };
272+		ssize_t len = 0;
273+		time_t rightnow = time (NULL);
274+		char *startrand = buf + sizeof (time_t);
275+		unsigned char *p = (unsigned char *) startrand;
276+		size_t randbytes = RANDOM_BYTES;
277+		int flags = O_RDONLY;
278+		int fd;
279+#ifdef O_NOCTTY
280+		flags |= O_NOCTTY;
281+#endif
282+		if (rightnow != (time_t)-1)
283+			while (rightnow > 0) {
284+				*--p = rightnow % (UCHAR_MAX + 1);
285+				rightnow /= UCHAR_MAX + 1;
286+			}
287+		else {
288+			/* try to use more random data */
289+			randbytes = COMMITID_RAW_SIZE;
290+			startrand = buf;
291+		}
292+		fd = open (urandom_dev, flags);
293+		if (fd >= 0) {
294+			len = read (fd, startrand, randbytes);
295+			close (fd);
296+		}
297+		if (len <= 0) {
298+			/* no random data was available so use pid */
299+			long int pid = (long int)getpid ();
300+			p = (unsigned char *) (startrand + sizeof (pid));
301+			while (pid > 0) {
302+			    *--p = pid % (UCHAR_MAX + 1);
303+			    pid /= UCHAR_MAX + 1;
304+			}
305+		}
306+		convert(buf, commitid);
307+	}
308+
309 	argc = getRCSINIT(argc, argv, &newargv);
310 	argv = newargv;
311 	while (a = *++argv,  0<--argc && *a++=='-') {
312@@ -532,6 +576,8 @@ mainProg(ciId, "ci", "$Id: ci.c,v 5.30 1
313 	newdelta.name = 0;
314 	clear_buf(&newdelta.ig);
315 	clear_buf(&newdelta.igtext);
316+	/* set commitid */
317+	newdelta.commitid=commitid;
318 	/* set author */
319 	if (author)
320 		newdelta.author=author;     /* set author given by -w         */
321@@ -1317,3 +1363,38 @@ addassoclst(flag, sp)
322 	*nextassoc = pt;
323 	nextassoc = &pt->nextsym;
324 }
325+
326+static char const alphabet[62] =
327+  "0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ";
328+
329+/* Divide BUF by D, returning the remainder.  Replace BUF by the
330+   quotient.  BUF[0] is the most significant part of BUF.
331+   D must not exceed UINT_MAX >> CHAR_BIT.  */
332+static unsigned int
333+divide_by (unsigned char buf[COMMITID_RAW_SIZE], unsigned int d)
334+{
335+  unsigned int carry = 0;
336+  int i;
337+  for (i = 0; i < COMMITID_RAW_SIZE; i++)
338+    {
339+      unsigned int byte = buf[i];
340+      unsigned int dividend = (carry << CHAR_BIT) + byte;
341+      buf[i] = dividend / d;
342+      carry = dividend % d;
343+    }
344+  return carry;
345+}
346+
347+static void
348+convert (char const input[COMMITID_RAW_SIZE], char *output)
349+{
350+  static char const zero[COMMITID_RAW_SIZE] = { 0, };
351+  unsigned char buf[COMMITID_RAW_SIZE];
352+  size_t o = 0;
353+  memcpy (buf, input, COMMITID_RAW_SIZE);
354+  while (memcmp (buf, zero, COMMITID_RAW_SIZE) != 0)
355+    output[o++] = alphabet[divide_by (buf, sizeof alphabet)];
356+  if (! o)
357+    output[o++] = '0';
358+  output[o] = '\0';
359+}
360Index:src/rcsgen.c
361--- src/rcsgen.c~	1995-06-16 06:19:24.000000000 +0000
362+++ src/rcsgen.c	2005-09-27 22:08:47.421504000 +0000
363@@ -547,6 +547,9 @@ putdelta(node, fout)
364
365 	aprintf(fout, ";\n%s\t%s;\n", Knext, node->next?node->next->num:"");
366 	awrite(node->ig.string, node->ig.size, fout);
367+
368+	if (node->commitid)
369+		aprintf(fout, "%s\t%s;\n", Kcommitid, node->commitid);
370 }
371
372
373Index:src/rcssyn.c
374--- src/rcssyn.c~	1995-06-16 06:19:24.000000000 +0000
375+++ src/rcssyn.c	2005-09-27 22:08:47.429504000 +0000
376@@ -171,6 +171,7 @@ char const
377 	Kauthor[]   = "author",
378 	Kbranch[]   = "branch",
379 	Kcomment[]  = "comment",
380+	Kcommitid[] = "commitid",
381 	Kdate[]     = "date",
382 	Kdesc[]     = "desc",
383 	Kexpand[]   = "expand",
384@@ -433,6 +434,13 @@ getdelta()
385 	Delta->lockedby = 0;
386 	Delta->log.string = 0;
387 	Delta->selector = true;
388+
389+	if (getkeyopt(Kcommitid)) {
390+		Delta->commitid = NextString;
391+		nextlex();
392+		getsemi(Kcommitid);
393+        }
394+
395 	Delta->ig = getphrases(Kdesc);
396         TotalDeltas++;
397         return (true);
398Index:src/rlog.c
399--- src/rlog.c~	1995-06-16 06:19:24.000000000 +0000
400+++ src/rlog.c	2005-09-26 17:23:55.257504000 +0000
401@@ -591,6 +591,10 @@ putadelta(node,editscript,trunk)
402 	      aprintf(out, insDelFormat,
403                              editscript->insertlns, editscript->deletelns);
404
405+	if ( node->commitid )
406+	   aprintf(out, "%s commitid: %s", (editscript) ? ";" : "",
407+		   node->commitid);
408+
409         newbranch = node->branches;
410         if ( newbranch ) {
411 	   bufautobegin(&branchnum);
412