FYI: Sed portability (Was: sed usage in autoheader)

[Akim Demaille]

| > From: Akim Demaille <address@hidden>
| > Date: 15 Nov 2000 16:32:18 +0100
| > 
| > Do you mean `s,[^/]*$,,' would work here?  Man, how can someone write
| > something that broken :(.
| 
| Just for the record, Cray is correct: POSIX does not allow
| `s/[^/]*$//'.  `s,[^/]*$,,' is required to work, though.

I'm applying the path below, thanks!
 
| > Bruce> never, ever use substitution separator
| > Bruce> chars inside of the substitution expressions, period.
| 
| I've had good luck with strings like `/foo[\/]bar/' in cases where I
| already know from context that the matched string can't contain `\'.
| That's portable, as it works with both the traditional and the POSIX
| interpretation.

We do agree `/foo[\/]bar/' is a single pattern which matches the
string `foo/bar', right?  Then, why do you use the char class here?

Index: ChangeLog
from  Akim Demaille  <address@hidden>

        POSIX doesn't require s/[^/]// to work.
        From Paul Eggert and Johan Danielsson.

        * doc/autoconf.texi (Limitations of Usual Tools) <sed>:
        Reorganize.  Document this issue.
        * autoupdate.sh (dir): Use `,' as separator instead of `/'
        * autoreconf.sh: Likewise.
        * autoupdate.sh: Ditto.

Index: autoheader.sh
===================================================================
RCS file: /cvs/autoconf/autoheader.sh,v
retrieving revision 1.87
diff -u -u -r1.87 autoheader.sh
--- autoheader.sh 2000/11/07 17:58:43 1.87
+++ autoheader.sh 2000/11/16 09:49:42
@@ -81,7 +81,7 @@
 
 # Variables.
 : address@hidden@}}
-dir=`echo "$0" | sed -e 's/[^/]*$//'`
+dir=`echo "$0" | sed -e 's,[^/]*$,,'`
 # We test "$dir/autoconf" in case we are in the build tree, in which case
 # the names are not transformed yet.
 for autoconf in "$AUTOCONF" \
@@ -293,7 +293,7 @@
 
 # Dump `acconfig.h' but its bottom.
 test -r $localdir/acconfig.h &&
-  sed -e '/@BOTTOM@/,$d' -e 's/@TOP@//' $localdir/acconfig.h >>$tmp/config.hin
+  sed '/@BOTTOM@/,$d;s/@TOP@//' $localdir/acconfig.h >>$tmp/config.hin
 
 # Dump the templates from `configure.in'.
 for verb in `(set) 2>&1 | sed -n -e '/^ac_verbatim/s/^\([^=]*\)=.*$/\1/p' | 
sort`; do
Index: autoreconf.sh
===================================================================
RCS file: /cvs/autoconf/autoreconf.sh,v
retrieving revision 1.58
diff -u -u -r1.58 autoreconf.sh
--- autoreconf.sh 2000/11/07 17:58:43 1.58
+++ autoreconf.sh 2000/11/16 09:49:42
@@ -91,7 +91,7 @@
 # Variables.
 : address@hidden@}}
 debug=false
-dir=`echo "$0" | sed -e 's/[^/]*$//'`
+dir=`echo "$0" | sed -e 's,[^/]*$,,'`
 force=false
 # --install -- as --add-missing in other tools.
 install=false
Index: autoupdate.sh
===================================================================
RCS file: /cvs/autoconf/autoupdate.sh,v
retrieving revision 1.44
diff -u -u -r1.44 autoupdate.sh
--- autoupdate.sh 2000/11/07 17:58:43 1.44
+++ autoupdate.sh 2000/11/16 09:49:50
@@ -96,7 +96,7 @@
 # Variables.
 : address@hidden@}}
 : ${SIMPLE_BACKUP_SUFFIX='~'}
-dir=`echo "$0" | sed 's/[^/]*$//'`
+dir=`echo "$0" | sed 's,[^/]*$,,'`
 # We test "$dir/autoconf" in case we are in the build tree, in which case
 # the names are not transformed yet.
 for autoconf in "$AUTOCONF" \
Index: doc/autoconf.texi
===================================================================
RCS file: /cvs/autoconf/doc/autoconf.texi,v
retrieving revision 1.388
diff -u -u -r1.388 autoconf.texi
--- doc/autoconf.texi 2000/11/14 16:01:05 1.388
+++ doc/autoconf.texi 2000/11/16 09:54:11
@@ -5885,8 +5885,26 @@
 
 @item @command{sed}
 @cindex @command{sed}
+Patterns should not include the separator (unless escaped), even as part
+of a character class.  In conformance with @sc{posix}, the Cray
address@hidden will reject @samp{s/[^/]*$//}: use @samp{s,[^/]*$,,}.
+
+Sed scripts should not use branch labels longer than 8 characters and
+should not contain comments.
+
+Input should have reasonably long lines, since some @command{sed} have
+an input buffer limited to 4000 bytes.
+
+Alternation, @samp{\|}, is common but not portable.
address@hidden FIXME: I know Solaris is guilty, but I don't remember how.
+Anchors (@samp{^} and @samp{$}) inside groups are not portable.
+
+Nested groups are extremely portable, but there is at least one
address@hidden (System V/68 Base Operating System R3V7.1) that does not
+support it.
+
 Of course the option @option{-e} is portable, but it is not needed.  No
-valid @command{sed} program can start with a dash, so it does not help
+valid Sed program can start with a dash, so it does not help
 disambiguating.  Its sole usefulness is helping enforcing indenting as
 in:
 
@@ -5905,21 +5923,6 @@
 Contrary to yet another urban legend, you may portably use @samp{&} in
 the replacement part of the @code{s} command to mean ``what was
 matched''.
-
address@hidden scripts should not use branch labels longer than 8
-characters and should not contain comments.
-
address@hidden input should have reasonably long lines, since some
address@hidden have an input buffer limited to 4000 bytes.
-
-Alternation, @samp{\|}, is common but not portable.
-
-Nested groups are extremely portable, but there is at least one
address@hidden (System V/68 Base Operating System R3V7.1) that does not
-support it.
-
-Anchors (@samp{^} and @samp{$}) inside groups are not portable.
address@hidden FIXME: I know Solaris is guilty, but I don't remember how.
 
 
 @item @command{sed} (@samp{t})