PATCH: installing executable functions

PATCH: installing executable functions

[Peter Stephenson]

On the subject of "#!" lines, and given the general flakiness of using
the executable bit on files from CVS, wouldn't it be better and
generally more consistent to make installed functions executable if and
only if they have a #! line?  If this is doing the wrong thing with a
function, then presumably the #! line is misplaced anyway.

("sed -ne '1p'" is in case there's some ghastliness where "head -1" is
broken as has occasionally been observed with "tail -1".)

The biggest disadvantage here is this could make function installation
on some systems, notably Cygwin, even more interminable.  Unfortunately
we're relying on sh at this point.

Index: Config/installfns.sh
===================================================================
RCS file: /cvsroot/zsh/zsh/Config/installfns.sh,v
retrieving revision 1.14
diff -u -r1.14 installfns.sh
--- Config/installfns.sh	26 Jul 2007 08:58:08 -0000	1.14
+++ Config/installfns.sh	29 Jan 2010 21:49:56 -0000
@@ -46,7 +46,7 @@
     fi
     test -d $instdir || /bin/sh $sdir_top/mkinstalldirs $instdir || exit 1
     $INSTALL_DATA $sdir_top/$file $instdir || exit 1
-    if test -x $sdir_top/$file; then
+    if sed -ne '1p' $sdir_top/$file | grep '^#!' >/dev/null; then
 	chmod +x $instdir/`echo $file | sed -e 's%^.*/%%'`
     fi
   fi

-- 
Peter Stephenson <p.w.stephenson@ntlworld.com>
Web page now at http://homepage.ntlworld.com/p.w.stephenson/

Re: PATCH: installing executable functions

[Bart Schaefer]

On Jan 29,  9:54pm, Peter Stephenson wrote:
}
} ("sed -ne '1p'" is in case there's some ghastliness where "head -1" is
} broken as has occasionally been observed with "tail -1".)
} 
} The biggest disadvantage here is this could make function installation
} on some systems, notably Cygwin, even more interminable.

Perhaps this would be faster?

  if test -n "`sed -ne 2q -e '/^#!/p' $sdir_top/$file`"; then

} +    if sed -ne '1p' $sdir_top/$file | grep '^#!' >/dev/null; then

Re: PATCH: installing executable functions

[Andrey Borzenkov]

[-- Attachment #1: Type: Text/Plain, Size: 747 bytes --]

On Saturday 30 of January 2010 03:36:45 Bart Schaefer wrote:
> On Jan 29,  9:54pm, Peter Stephenson wrote:
> }
> } ("sed -ne '1p'" is in case there's some ghastliness where "head -1"
> is } broken as has occasionally been observed with "tail -1".)
> }
> } The biggest disadvantage here is this could make function
> installation } on some systems, notably Cygwin, even more
> interminable.
> 
> Perhaps this would be faster?
> 
>   if test -n "`sed -ne 2q -e '/^#!/p' $sdir_top/$file`"; then
> 
> } +    if sed -ne '1p' $sdir_top/$file | grep '^#!' >/dev/null; then

Hmm ... to completely avoid fork?

read -r line < $sdir_top/$file
case "$line" in
    '#!'* )
		chmod +x $instdir/`echo $file | sed -e 's%^.*/%%'`
	;;
esac

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: PATCH: installing executable functions

[Peter Stephenson]

On Sat, 30 Jan 2010 08:35:03 +0300
Andrey Borzenkov <arvidjaar@gmail.com> wrote:
> read -r line < $sdir_top/$file
> case "$line" in
>     '#!'* )
> 		chmod +x $instdir/`echo $file | sed -e 's%^.*/%%'`
> 	;;
> esac

This looks pretty good: we already rely on "case" heavily in
configure...  I'm a trifle worried about "read -r" in case some perverse
Bourne shell doesn't handle the "-r", and since we don't want spaces or
backslashes etc.  I don't think it's necessary... any theories / further
suggestions?

-- 
Peter Stephenson <p.w.stephenson@ntlworld.com>
Web page now at http://homepage.ntlworld.com/p.w.stephenson/