10:02 p.m.
I tried configuring 21.5.28 on Solaris10/Intel with this command:
LDFLAGS='-R/opt/app/nonc++/zlib-1.2/lib -L/opt/app/nonc++/zlib-1.2/lib \
-R/opt/app/nonc++/libpng-1.2/lib -L/opt/app/nonc++/libpng-1.2/lib \
-R/opt/app/nonc++/xpm-3.4k/lib -L/opt/app/nonc++/xpm-3.4k/lib' \
CPPFLAGS='-I/opt/app/nonc++/zlib-1.2/include \
-I/opt/app/nonc++/libpng-1.2/include -I/opt/app/nonc++/xpm-3.4k/include' \
./configure --prefix=/opt/app/xemacs-21.5.28 --with-optimization \
--with-error-checking=none
and get this error:
./configure: Usage error:
Invalid type list supplied. Valid types for the --with-error-checking
option are:
byte_code display extents gc glyphs malloc structures text types.
A prefix of "no" (eg, "notype" for "type") turns a type
off.
Types not explicitly specified or negated take their defaults.
The list of defaults is
"byte_code,display,extents,gc,glyphs,malloc,structures,text,types".
Prefixing `all' or
`none' to the list sets all defaults to on and off respectively.
`yes' as the sole argument is equivalent to omitting the argument.
Use `./configure --help' to show usage.
Using --help doesn't mention "none" or "all" either.
I worked around it with this setting:
--with-error-checking=nobyte_code,nodisplay,noextents,nogc,noglyphs,nomalloc,nostructures,notext,notypes
--
Skip Montanaro - skip(a)pobox.com - http://www.webfast.com/~skip/
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
7:10 a.m.
skip(a)pobox.com writes:
I tried configuring 21.5.28 on Solaris10/Intel with this command:
It doesn't reproduce on Linux. :-( With the commands "../configure
--with-optimization --with-error-checking=none" and "../configure
--with-error-checking=none", configure runs, and produces a
src/config.h with all error-checking disabled. (Using
--with-error-checking=all also runs and the config.h is correct.)
Maybe we managed to give you an old, broken configure. Do you have
autoconf available? If so could you try running autoconf, and
rerunning the configure command?
What shell are you using?
Did this configure script come from CVS or tarballs?
Use `./configure --help' to show usage.
Using --help doesn't mention "none" or "all" either.
OK, that's a bug I can fix.
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
1:06 p.m.
Stephen> Maybe we managed to give you an old, broken configure. Do you
Stephen> have autoconf available? If so could you try running autoconf,
Stephen> and rerunning the configure command?
Stephen> What shell are you using?
Stephen> Did this configure script come from CVS or tarballs?
That confused me me because it also worked fine out of the box on my Mac.
Rerunning autoconf (2.59) didn't help. I'm using bash 3.00.16(1)-release on
Solaris, 2.05b.0(1)-release on my Mac.
Okay, I tracked it down to what I think is a misuse of GNU sed features in
the configure script. On my Mac:
% with_error_checking=none
% echo $with_error_checking | sed -e 's/^[a-z]*\(,\(.*\)\)\{0,1\}$/\2/'
On the Solaris system at work:
% with_error_checking=none
% echo $with_error_checking | sed -e 's/^[a-z]*\(,\(.*\)\)\{0,1\}$/\2/'
none
Maybe configure (or autoconf) is expecting too much from Sun's sed. I don't
believe older versions of sed understand the {0,1} construct do they? I
think that's a GNU sed invention. If I replace that sed command with gsed
(our local install of GNU sed), it works.
Skip
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
1:17 p.m.
On 7/12/07, skip(a)pobox.com <skip(a)pobox.com> wrote:
Stephen> Maybe we managed to give you an old, broken configure. Do
you
Stephen> have autoconf available? If so could you try running
autoconf,
Stephen> and rerunning the configure command?
Stephen> What shell are you using?
Stephen> Did this configure script come from CVS or tarballs?
That confused me me because it also worked fine out of the box on my Mac.
Rerunning autoconf (2.59) didn't help. I'm using bash 3.00.16(1)-release
on
Solaris, 2.05b.0(1)-release on my Mac.
Okay, I tracked it down to what I think is a misuse of GNU sed features in
the configure script. On my Mac:
% with_error_checking=none
% echo $with_error_checking | sed -e
's/^[a-z]*\(,\(.*\)\)\{0,1\}$/\2/'
On the Solaris system at work:
% with_error_checking=none
% echo $with_error_checking | sed -e
's/^[a-z]*\(,\(.*\)\)\{0,1\}$/\2/'
none
Maybe configure (or autoconf) is expecting too much from Sun's sed. I
don't
believe older versions of sed understand the {0,1} construct do they? I
think that's a GNU sed invention. If I replace that sed command with gsed
(our local install of GNU sed), it works.
Just out of interest, does the sed in /usr/xpg4 work correctly? The tools
there tend to be more modern on Solaris.
Robert
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
1:55 p.m.
Robert> Just out of interest, does the sed in /usr/xpg4 work correctly?
Yes, it does. Thanks. Adding a symlink to that version of sed from an
earlier directory in PATH solves my configure problem (but see my next
mail...). I didn't want to insert /usr/xpg4/bin at the start of PATH for
fear of changing something else.
Just for completeness, /usr/ucb/sed does not work.
Skip
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
4:26 a.m.
skip(a)pobox.com writes:
Robert> Just out of interest, does the sed in /usr/xpg4 work correctly?
Yes, it does. Thanks. Adding a symlink to that version of sed from an
earlier directory in PATH solves my configure problem (but see my next
mail...). I didn't want to insert /usr/xpg4/bin at the start of PATH for
fear of changing something else.
Just for completeness, /usr/ucb/sed does not work.
Thank you both very much for this analysis.
I suppose that the '?' construct is understood by basically all seds?
Using '\{0,1\}' instead of '?' seems like asking for trouble....
I'll check the autoconf portability docs and see what they say, but if
anyone has any random information on sed script portability I'd love
to hear it.
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
4:40 a.m.
Stephen> I suppose that the '?' construct is understood by basically all
Stephen> seds? Using '\{0,1\}' instead of '?' seems like asking
for
Stephen> trouble....
Switching to ? doesn't seem to help:
% echo $with_error_checking | /usr/bin/sed -e
's/^[a-z]*\(,\(.*\)\)\{0,1\}$/\2/'
none
% echo $with_error_checking | /usr/bin/sed -e 's/^[a-z]*\(,\(.*\)\)?$/\2/'
none
% echo $with_error_checking | /usr/bin/sed -e 's/^[a-z]*\(,\(.*\)\)\?$/\2/'
none
Also fails with /usr/xpg4 version of sed:
% echo $with_error_checking | /usr/xpg4/bin/sed -e
's/^[a-z]*\(,\(.*\)\)?$/\2/'
none
Feel free to suggest other alternatives for me to try.
Skip
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
5:13 a.m.
skip> Feel free to suggest other alternatives for me to try.
Thinking about it a bit, shouldn't this work and be portable?
% echo $with_error_checking | /usr/bin/sed -e 's/^[a-z]*//' -e
's/^,//'
It gives the correct result in the current situation. If I am interpreting
the original r.e. correctly, the aim is to remove the first word from the
beginning of the string plus any comma that follows it.
Skip
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
9:44 a.m.
skip(a)pobox.com wrote:
skip> Feel free to suggest other alternatives for me to try.
Thinking about it a bit, shouldn't this work and be portable?
% echo $with_error_checking | /usr/bin/sed -e 's/^[a-z]*//' -e
's/^,//'
It gives the correct result in the current situation. If I am interpreting
the original r.e. correctly, the aim is to remove the first word from the
beginning of the string plus any comma that follows it.
Actually, what also works in all seds is:
/usr/bin/sed -e 's/^[a-z]*,\{0,1\}//'
Solaris /usr/bin/sed only supports so-called simple regexps, they
are explained in regexp(5). There is documented that the \{m,n\}
construct only works for preceding _single-character_ REs, but not
for parenthized REs.
Joachim
--
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Joachim Schrod Email: jschrod(a)acm.org
Roedermark, Germany
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
3:56 p.m.
Stephen J. Turnbull wrote:
> Robert> Just out of interest, does the sed in /usr/xpg4
work correctly?
>
> Yes, it does. Thanks. Adding a symlink to that version of sed from an
> earlier directory in PATH solves my configure problem (but see my next
> mail...). I didn't want to insert /usr/xpg4/bin at the start of PATH for
> fear of changing something else.
>
> Just for completeness, /usr/ucb/sed does not work.
Thank you both very much for this analysis.
I suppose that the '?' construct is understood by basically all seds?
Using '\{0,1\}' instead of '?' seems like asking for trouble....
I'll check the autoconf portability docs and see what they say, but if
anyone has any random information on sed script portability I'd love
to hear it.
According to IEEE 1003.1-2004[1], \{...\} is standard but ? or \?
aren't.
[1] http://www.opengroup.org/onlinepubs/009695399/utilities/sed.html
The sed utility shall support the BREs described in the Base
Definitions volume of IEEE Std 1003.1-2001, Section 9.3, Basic
Regular Expressions[2], with the following additions:
[2] http://www.opengroup.org/onlinepubs/009695399/basedefs/xbd_chap09.html#ta...
The Info file for GNU sed says that \? is a GNU extension.
--
Glynn Clements <glynn(a)gclements.plus.com>
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
4:29 p.m.
Glynn> According to IEEE 1003.1-2004[1], \{...\} is standard but ? or \?
Glynn> aren't.
Unlike many of its other man pages Sun's sed man page doesn't claim
conformance with any standards. I'm pretty sure it predates all these
fancy-schmancy standards documents.
Skip
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
8:11 p.m.
skip(a)pobox.com wrote:
Glynn> According to IEEE 1003.1-2004[1], \{...\} is standard
but ? or \?
Glynn> aren't.
Unlike many of its other man pages Sun's sed man page doesn't claim
conformance with any standards. I'm pretty sure it predates all these
fancy-schmancy standards documents.
Skip
The sed man page references XPG4 for, presumably, /usr/xpg4/bin/sed.
man XPG4 says Solaris 7-9 support SUSv2 (POSIX.1b-1993 and
POSIX.1c-1996) while Solaris 10 supports SUSv3 (POSIX.1-2001). This
should happen automatically if your environment has the setting
CONFIG_SHELL=/usr/xpg4/bin/sh as described in the PROBLEMS file.
Rodney
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
6343
days inactive
6345
days old
11 comments
6 participants
participants (6)
-
Glynn Clements -
Joachim Schrod -
Robert Pluim -
Rodney Sparapani -
Skip Montanaro -
Stephen J. Turnbull