On 03/10/11 20:38, Nigel Taylor wrote:
> On 03/10/11 16:28, Stuart Henderson wrote:
>> good find.
>>
>> after reading posix 2008 on this it isn't clear to me what is
>> specified, but GNU m4 is clear in the documentation that _they_
>> apply it non-recursively.
>>
>> http://pubs.opengroup.org/onlinepubs/9699919799/utilities/m4.html
>> http://www.gnu.org/software/m4/manual/html_node/Translit.html
>>
>> so imo we definitely want this for -g mode and need to consider
>> carefully whether to do it always (in which case the comment above
>> the start of map(), which explains why this is done, would also
>> need adjusting).
>>
>> the following from MirOS may also be of interest:
>>
>> http://junkpile.org/14.patch "fix trace lineno output for 'macro\n'"
>> http://junkpile.org/15.patch "let 'errprint' in 'm4 -g' mode behave like GNU"
>> http://junkpile.org/16.patch "fix another line number problem"
>>
>> i'll try and have a play with this and hopefully Marc will
>> have some time to look at it soon.
>>
>>
> Hi,
>
> xenocara build completed, no issue introduced there, also rebuilt a number of
> ports.
>
> Checking the info m4 on OpenBSD looks to be the same as gnu m4 manual page for
> translit. According to the last example does not do recursion
>
> "translit(`abcdef', `aabdef', `bcged')
>
> The final example shows that `a' is mapped to `b', not `c'; the resulting `b'
> is
> not further remapped to `g'; the `d' and `e' are swapped, and the `f' is
> discarded."
>
>
> "translit(`+,-12345', `+--1-5', `<;>a-c-a')
>
> This fails to give the correct reasult on OpenBSD a-c-a expected to be
> equivalent to abcba, -a on the end is treated as -a." Back to back ranges not
> implemented yet.
>
> I think the info m4 examples should be in the regression tests. All but the
> back
> to back range worked (-g only gnu), with the updated version attached, I have
> fixed that as below, result from test m4 vs gm4 as per m4 info.
>
> $ sh test_translit.sh
> [HAVE_abc def h/] [HAVE_abc def h/]
> [HAVE_abc~def~h/] [HAVE_abc~def~h/]
> [HAVE_abc/def~h/] [HAVE_abc/def~h/]
> [HAVE_abc/def;h/] [HAVE_abc/def;h/]
> [HAVE_ABC/def;h/] [HAVE_ABC/def;h/]
> [HAVE_ABCZdef;hZ] [HAVE_ABCZdef;hZ]
> ABCDEFGHIJ ABCDEFGHIJ
> ABCDEFGHIJ ABCDEFGHIJ
> ABC-0980-ZYX ABC-0980-ZYX
> ABC-0980-ZYX ABC-0980-ZYX
> s not nix s not nix
> GNUS NOT UNIX GNUS NOT UNIX
> tmfs not fnix tmfs not fnix
> <;>abcba <;>abcba
> bgced bgced
>
> I will have a look at the patches.
>
> Regards
>
> Nigel Taylor
>
>
>
>
Hi,
I missed incrementing pointer, after back-to-back range.
Regards
Nigel Taylor
Index: usr.bin/m4/eval.c
===================================================================
RCS file: /home/cvs/src/usr.bin/m4/eval.c,v
retrieving revision 1.68
diff -u -p -r1.68 eval.c
--- usr.bin/m4/eval.c 7 Sep 2010 19:58:09 -0000 1.68
+++ usr.bin/m4/eval.c 10 Mar 2011 20:58:34 -0000
@@ -884,21 +884,11 @@ dosub(const char *argv[], int argc)
* language. Within mapvec, we replace every character of "from" with
* the corresponding character in "to". If "to" is shorter than "from",
* than the corresponding entries are null, which means that those
- * characters dissapear altogether. Furthermore, imagine
- * map(dest, "sourcestring", "srtin", "rn..*") type call. In this case,
- * `s' maps to `r', `r' maps to `n' and `n' maps to `*'. Thus, `s'
- * ultimately maps to `*'. In order to achieve this effect in an efficient
- * manner (i.e. without multiple passes over the destination string), we
- * loop over mapvec, starting with the initial source character. if the
- * character value (dch) in this location is different than the source
- * character (sch), sch becomes dch, once again to index into mapvec, until
- * the character value stabilizes (i.e. sch = dch, in other words
- * mapvec[n] == n). Even if the entry in the mapvec is null for an ordinary
- * character, it will stabilize, since mapvec[0] == 0 at all times. At the
- * end, we restore mapvec* back to normal where mapvec[n] == n for
- * 0 <= n <= 127. This strategy, along with the restoration of mapvec, is
- * about 5 times faster than any algorithm that makes multiple passes over
- * destination string.
+ * characters dissapear altogether.
+ * The recursion has been removed to match gnu m4 implementation and
+ * matches the m4 info details.
+ * At the end, we restore mapvec* back to normal where mapvec[n] == n for
+ * 0 <= n <= 255.
*/
static void
map(char *dest, const char *src, const char *from, const char *to)
@@ -958,10 +948,6 @@ map(char *dest, const char *src, const c
while (*src) {
sch = (unsigned char)(*src++);
dch = mapvec[sch];
- while (dch != sch) {
- sch = dch;
- dch = mapvec[sch];
- }
if ((*dest = (char)dch))
dest++;
}
@@ -993,7 +979,7 @@ handledash(char *buffer, char *end, cons
unsigned char i;
if ((unsigned char)src[0] <= (unsigned char)src[2]) {
for (i = (unsigned char)src[0];
- i <= (unsigned char)src[2]; i++) {
+ i < (unsigned char)src[2]; i++) {
*p++ = i;
if (p == end) {
*p = '\0';
@@ -1002,7 +988,7 @@ handledash(char *buffer, char *end, cons
}
} else {
for (i = (unsigned char)src[0];
- i >= (unsigned char)src[2]; i--) {
+ i > (unsigned char)src[2]; i--) {
*p++ = i;
if (p == end) {
*p = '\0';
@@ -1010,7 +996,37 @@ handledash(char *buffer, char *end, cons
}
}
}
- src += 3;
+ src += 2;
+/* check for back to back range */
+ if (src[1] == '-' && src[2]) {
+ if ((unsigned char)src[0] <= (unsigned
char)src[2]) {
+ for (i = (unsigned char)src[0];
+ i <= (unsigned char)src[2]; i++) {
+ *p++ = i;
+ if (p == end) {
+ *p = '\0';
+ return buffer;
+ }
+ }
+ } else {
+ for (i = (unsigned char)src[0];
+ i >= (unsigned char)src[2]; i--) {
+ *p++ = i;
+ if (p == end) {
+ *p = '\0';
+ return buffer;
+ }
+ }
+ }
+ src += 3;
+ }
+ else {
+ *p++ = *src++;
+ if (p == end) {
+ *p = '\0';
+ return buffer;
+ }
+ }
} else
*p++ = *src++;
if (p == end)
Index: regress/usr.bin/m4/Makefile
===================================================================
RCS file: /home/cvs/src/regress/usr.bin/m4/Makefile,v
retrieving revision 1.28
diff -u -p -r1.28 Makefile
--- regress/usr.bin/m4/Makefile 23 Mar 2010 20:11:52 -0000 1.28
+++ regress/usr.bin/m4/Makefile 10 Mar 2011 14:22:45 -0000
@@ -12,7 +12,7 @@ REGRESS_TARGETS= test-ff_after_dnl test-
test-weird test-args test-args2 test-esyscmd test-eval test-gnupatterns \
test-gnupatterns2 test-comments test-synch1 test-synch1bis \
test-gnuformat test-includes test-dumpdef test-gnuprefix \
- test-translit
+ test-translit test-gnutranslit
test-ff_after_dnl: ff_after_dnl.m4
${M4} ff_after_dnl.m4 | diff - ${.CURDIR}/ff_after_dnl.out
@@ -102,6 +102,9 @@ test-dumpdef:
test-gnuprefix:
${M4} -P ${.CURDIR}/gnuprefix.m4 2>&1 | \
diff -u - ${.CURDIR}/gnuprefix.out
+
+test-gnutranslit:
+ ${M4} -g ${.CURDIR}/gnutranslit.m4 | diff -u -
${.CURDIR}/gnutranslit.out
.PHONY: ${REGRESS_TARGETS}
Index: regress/usr.bin/m4/gnutranslit.m4
===================================================================
RCS file: regress/usr.bin/m4/gnutranslit.m4
diff -N regress/usr.bin/m4/gnutranslit.m4
--- /dev/null 1 Jan 1970 00:00:00 -0000
+++ regress/usr.bin/m4/gnutranslit.m4 10 Mar 2011 20:31:20 -0000
@@ -0,0 +1,13 @@
+translit(`[HAVE_abc/def.h
+]', `
+/.', `/ ')
+translit(`[HAVE_abc/def.h=]', `=/.', `/~~')
+translit(`0123456789', `0123456789', `ABCDEFGHIJ')
+translit(`0123456789', `[0-9]', `[A-J]')
+translit(`abc-0980-zyx', `abcdefghijklmnopqrstuvwxyz',
`ABCDEFGHIJKLMNOPQRSTUVWXYZ')
+translit(`abc-0980-zyx', `[a-z]', `[A-Z]')
+translit(`GNUs not Unix', `A-Z')
+translit(`GNUs not Unix', `a-z', `A-Z')
+translit(`GNUs not Unix', `A-Z', `z-a')
+translit(`+,-12345', `+--1-5', `<;>a-c-a')
+translit(`abcdef', `aabdef', `bcged')
Index: regress/usr.bin/m4/gnutranslit.out
===================================================================
RCS file: regress/usr.bin/m4/gnutranslit.out
diff -N regress/usr.bin/m4/gnutranslit.out
--- /dev/null 1 Jan 1970 00:00:00 -0000
+++ regress/usr.bin/m4/gnutranslit.out 10 Mar 2011 20:31:28 -0000
@@ -0,0 +1,11 @@
+[HAVE_abc def h/]
+[HAVE_abc~def~h/]
+ABCDEFGHIJ
+ABCDEFGHIJ
+ABC-0980-ZYX
+ABC-0980-ZYX
+s not nix
+GNUS NOT UNIX
+tmfs not fnix
+<;>abcba
+bgced
