Bug#224301: [Adduser-devel] Bug#224301: Some thoughts about this bug report
On Tue, Oct 25, 2005 at 07:21:59AM +0200, Marc Haber wrote:
Frankly, I see no reason to keep this wrapper. If you do not want
localized messages, run your scripts with LC_ALL=C.
It is consistent with the way gettext is used in C programs,
I tend to disagree, but that's really not important ;)
and it allows gettext to be disabled completely, which might be an
advantage on low memory systems.
I do not know how to perform reliable memory benchmarks, but am pretty
sure that disabling gettext has no noticeable impact because localized
messages are not used at all with LC_ALL=C:
$ strace -e open /usr/sbin/adduser --help 21 /dev/null | grep locale
...
open(/usr/share/locale/fr_FR/LC_MESSAGES/adduser.mo, O_RDONLY) = -1 ENOENT
(No such file or directory)
open(/usr/share/locale/fr/LC_MESSAGES/adduser.mo, O_RDONLY) = 3
$ LC_ALL=C strace -e open /usr/sbin/adduser --help 21 /dev/null | grep
locale
$
In GNU libc, gettext() returns its argument in this case (and also when
no msgstr was found), so there is no string copy. If Locale::gettext
does the same (no idea whether this is true or not), gettext (with
LC_ALL=C) is very similar to
*gettext = sub { shift };
I have committed your changes and will upload 3.78 later today. Can
you then please take a new look at the package?
Sure. At the moment it did not yet reach my mirror, will look at it
later. Thanks for taking care.
Denis
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#224301: [Adduser-devel] Bug#224301: Some thoughts about this bug report
On Sat, Oct 22, 2005 at 09:02:28PM +0200, Marc Haber wrote:
Of course this comment may be improved ;)
You need to call xgettext with the -c flag so that comments are
inserted
into PO files.
that would be XGETTEXT = xgettext -c in po/Makefile, right? Done.
Hmmm, I prefer adding -c after $(XGETTEXT).
There is another glitch with po/Makefile:
adduser-3.76/po$ touch ../adduser
adduser-3.76/po$ make adduser.pot
xgettext -c -L Perl -kgtx -o adduser.pot ../adduser
Strings from deluser are then discarded. Patch attached.
Other minor i18n problems with your package:
* the gtx() function is useless and confuses xgettext
gtx was named _ previously and is obviously modeled after the _ macro
suggested in the gettext texinfo documentation. I'd like to keep it to
be able to disable i18n for testing and debugging purposes.
Unfortunately, it cannot be named _ as File::Find starts to behave
very strangely when a function called _ is in the namespace.
Is it possible to overload gettext with a no-op for testing purposes?
If so, your changes are acceptable. I'd like to have gettext in a
state where it can be disabled.
Can't xgettexts confusion be remedied by replacing -k_ with -kgtx in
the Makefile? I have tried this in the lastest upload to experimental,
please verify.
Yes, but xgettext still displays a warning.
In AdduserCommon.pm, you can replace
sub gtx {
return gettext( join , @_);
}
by
sub gtx {
return gettext(shift):
}
to remove this warning. Calling gtx with a list is wrong, because
only the first string is extracted into the POT file (see the attached
patch, the 2nd string from deluser is not in adduser.pot).
Here is a patch. I tested it, but of course it may introduce bugs,
please review it carefully ;)
I applied the changes, but left gtx intact. Please reconsider this
suggestion.
Frankly, I see no reason to keep this wrapper. If you do not want
localized messages, run your scripts with LC_ALL=C.
s_print (gtx(Backing up files to be removed to . $config{backup_to}.
...\n));
Doesn't this cause translation pain as well?
Absolutely, I missed this one.
I have changed it to
s_printf (gtx(Backing up files to be removed to %s
...\n),$config{backup_to});
Thanks.
Denis
diff -ur adduser-3.76.old/AdduserCommon.pm adduser-3.76/AdduserCommon.pm
--- adduser-3.76.old/AdduserCommon.pm 2005-10-18 16:20:00.0 +0200
+++ adduser-3.76/AdduserCommon.pm 2005-10-24 23:22:57.0 +0200
@@ -56,8 +56,8 @@
}
}
-sub gtx {
-return gettext( join , @_);
+sub gtx ($) {
+return gettext(shift);
}
sub dief {
diff -ur adduser-3.76.old/deluser adduser-3.76/deluser
--- adduser-3.76.old/deluser2005-10-22 20:54:53.0 +0200
+++ adduser-3.76/deluser2005-10-24 23:32:46.0 +0200
@@ -397,7 +397,7 @@
printf gtx(Copyright (C) 2000 Roland Bauerschmidt [EMAIL
PROTECTED]\n\n);
-printf gtx(deluser is based on adduser by Guy Maor [EMAIL PROTECTED],
Ian Murdock\n,
+printf gtx(deluser is based on adduser by Guy Maor [EMAIL PROTECTED],
Ian Murdock\n.
[EMAIL PROTECTED] and Ted Hajek [EMAIL PROTECTED]\n);
printf gtx(\nThis program is free software; you can redistribute it
and/or modify
diff -ur adduser-3.76.old/po/Makefile adduser-3.76/po/Makefile
--- adduser-3.76.old/po/Makefile2005-10-22 21:08:37.0 +0200
+++ adduser-3.76/po/Makefile2005-10-24 23:35:29.0 +0200
@@ -1,4 +1,4 @@
-XGETTEXT = xgettext -c
+XGETTEXT = xgettext
MSGFMT = msgfmt
MSGMERGE = msgmerge
@@ -12,6 +12,7 @@
PO = $(wildcard *.po)
LANG = $(basename $(PO))
MO = $(addsuffix .mo,$(LANG))
+SOURCES = ../adduser ../deluser ../AdduserCommon.pm
all: update $(MO)
update: adduser.pot
@@ -20,8 +21,8 @@
$(MSGMERGE) -U $$po adduser.pot; \
done;
-adduser.pot: ../adduser ../deluser ../AdduserCommon.pm
- $(XGETTEXT) -L Perl -kgtx -o $@ $?
+adduser.pot: $(SOURCES)
+ $(XGETTEXT) -c -L Perl -kgtx -o $@ $(SOURCES)
install: all
for i in $(MO) ; do \
Bug#224301: [Adduser-devel] Bug#224301: Some thoughts about this bug report
On Mon, Oct 24, 2005 at 11:51:56PM +0200, Denis Barbier wrote:
On Sat, Oct 22, 2005 at 09:02:28PM +0200, Marc Haber wrote:
Of course this comment may be improved ;)
You need to call xgettext with the -c flag so that comments are
inserted
into PO files.
that would be XGETTEXT = xgettext -c in po/Makefile, right? Done.
Hmmm, I prefer adding -c after $(XGETTEXT).
Changed.
There is another glitch with po/Makefile:
adduser-3.76/po$ touch ../adduser
adduser-3.76/po$ make adduser.pot
xgettext -c -L Perl -kgtx -o adduser.pot ../adduser
Strings from deluser are then discarded. Patch attached.
Applied.
Yes, but xgettext still displays a warning.
In AdduserCommon.pm, you can replace
sub gtx {
return gettext( join , @_);
}
by
sub gtx {
return gettext(shift):
}
to remove this warning. Calling gtx with a list is wrong, because
only the first string is extracted into the POT file (see the attached
patch, the 2nd string from deluser is not in adduser.pot).
Applied.
Here is a patch. I tested it, but of course it may introduce bugs,
please review it carefully ;)
I applied the changes, but left gtx intact. Please reconsider this
suggestion.
Frankly, I see no reason to keep this wrapper. If you do not want
localized messages, run your scripts with LC_ALL=C.
It is consistent with the way gettext is used in C programs, and it
allows gettext to be disabled completely, which might be an advantage
on low memory systems.
I have committed your changes and will upload 3.78 later today. Can
you then please take a new look at the package?
Greetings
Marc
--
-
Marc Haber | I don't trust Computers. They | Mailadresse im Header
Mannheim, Germany | lose things.Winona Ryder | Fon: *49 621 72739834
Nordisch by Nature | How to make an American Quilt | Fax: *49 621 72739835
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#224301: [Adduser-devel] Bug#224301: Some thoughts about this bug report
On Sat, Oct 22, 2005 at 12:02:43AM +0200, Denis Barbier wrote:
There is very good documentation in libc.info, section Yes-or-No Questions,
but unfortunately it is not relevant here because adduser is a perl script.
So I had a look at perllocale(1), which has some nice tips, in particular
about I18N::Langinfo. But be warned that their example is not very
valuable under GNU/Linux, because glibc introduced YESSTR/NOSTR lately
and many locales do not provide them. On the other hand, they all
provide YESEXPR/NOEXPR, so you can run something like:
use I18N::Langinfo qw(langinfo YESEXPR);
my $yesexpr = langinfo(YESEXPR());
for (;;) {
systemcall('/usr/bin/chfn', $new_name);
# Translators: [y/N] has to be replaced by values defined in your
# locale. You can see by running locale yesexpr which regular
# expression will be checked to find positive answer.
print (gtx(Is the information correct? [y/N] ));
chop ($answer=STDIN);
last if ($answer =~ m/$yesexpr/o);
}
Applied, thanks.
Of course this comment may be improved ;)
You need to call xgettext with the -c flag so that comments are inserted
into PO files.
that would be XGETTEXT = xgettext -c in po/Makefile, right? Done.
Other minor i18n problems with your package:
* the gtx() function is useless and confuses xgettext
gtx was named _ previously and is obviously modeled after the _ macro
suggested in the gettext texinfo documentation. I'd like to keep it to
be able to disable i18n for testing and debugging purposes.
Unfortunately, it cannot be named _ as File::Find starts to behave
very strangely when a function called _ is in the namespace.
Is it possible to overload gettext with a no-op for testing purposes?
If so, your changes are acceptable. I'd like to have gettext in a
state where it can be disabled.
Can't xgettexts confusion be remedied by replacing -k_ with -kgtx in
the Makefile? I have tried this in the lastest upload to experimental,
please verify.
Here is a patch. I tested it, but of course it may introduce bugs,
please review it carefully ;)
I applied the changes, but left gtx intact. Please reconsider this
suggestion.
s_print (gtx(Backing up files to be removed to . $config{backup_to}.
...\n));
Doesn't this cause translation pain as well?
I have changed it to
s_printf (gtx(Backing up files to be removed to %s
...\n),$config{backup_to});
--
-
Marc Haber | I don't trust Computers. They | Mailadresse im Header
Mannheim, Germany | lose things.Winona Ryder | Fon: *49 621 72739834
Nordisch by Nature | How to make an American Quilt | Fax: *49 621 72739835
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#224301: [Adduser-devel] Bug#224301: Some thoughts about this bug report
On Fri, Oct 21, 2005 at 08:36:10AM +0200, Christian Perrier wrote: (Marc pinged me on IRC about this bug report in adduser) Thanks for your help! First of all, I think that adduser should probably make better use of the locales information for this Yes/No answer. I think that would be a good idea here. I don't have the required skills to lead you to the solution, Marc, but I think that Denis Barbier could. Thanks for directly Ccing him, I hope that he can find the time to help. This fr.po file for adduser is under work now and you should soon have a new version anyway with these fixes among others. OK, I'll wait for that before applying stuff. Better not mess with things that I don't understand. Greetings Marc -- - Marc Haber | I don't trust Computers. They | Mailadresse im Header Mannheim, Germany | lose things.Winona Ryder | Fon: *49 621 72739834 Nordisch by Nature | How to make an American Quilt | Fax: *49 621 72739835 -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#224301: [Adduser-devel] Bug#224301: Some thoughts about this bug report
This fr.po file for adduser is under work now and you should soon have a new version anyway with these fixes among others. OK, I'll wait for that before applying stuff. Better not mess with things that I don't understand. You should also have an incoming mail by Thomas Huriaux (thomash on IRC) who mentioned me a problem in the package build system, as the package is built with svn-buildpackage, but re-synced PO files are not re-commited to your SVN...which is thus always outdated..:-) (at least, this is Thomas analysis) -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#224301: [Adduser-devel] Bug#224301: Some thoughts about this bug report
On Fri, Oct 21, 2005 at 06:03:34PM +0200, Christian Perrier wrote: This fr.po file for adduser is under work now and you should soon have a new version anyway with these fixes among others. OK, I'll wait for that before applying stuff. Better not mess with things that I don't understand. You should also have an incoming mail by Thomas Huriaux (thomash on IRC) who mentioned me a problem in the package build system, as the package is built with svn-buildpackage, but re-synced PO files are not re-commited to your SVN...which is thus always outdated..:-) (at least, this is Thomas analysis) I need to learn more about these processes, what gets done at which phase of package building. And obviously, there is no easy document available yet. Greetings Marc -- - Marc Haber | I don't trust Computers. They | Mailadresse im Header Mannheim, Germany | lose things.Winona Ryder | Fon: *49 621 72739834 Nordisch by Nature | How to make an American Quilt | Fax: *49 621 72739835 -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#224301: [Adduser-devel] Bug#224301: Some thoughts about this bug report
On Fri, Oct 21, 2005 at 09:23:36AM +0200, Marc Haber wrote:
On Fri, Oct 21, 2005 at 08:36:10AM +0200, Christian Perrier wrote:
(Marc pinged me on IRC about this bug report in adduser)
Thanks for your help!
First of all, I think that adduser should probably make better use of
the locales information for this Yes/No answer.
I think that would be a good idea here.
I don't have the required skills to lead you to the solution, Marc,
but I think that Denis Barbier could.
Thanks for directly Ccing him, I hope that he can find the time to help.
This fr.po file for adduser is under work now and you should soon have
a new version anyway with these fixes among others.
OK, I'll wait for that before applying stuff. Better not mess with
things that I don't understand.
Hi,
There is very good documentation in libc.info, section Yes-or-No Questions,
but unfortunately it is not relevant here because adduser is a perl script.
So I had a look at perllocale(1), which has some nice tips, in particular
about I18N::Langinfo. But be warned that their example is not very
valuable under GNU/Linux, because glibc introduced YESSTR/NOSTR lately
and many locales do not provide them. On the other hand, they all
provide YESEXPR/NOEXPR, so you can run something like:
use I18N::Langinfo qw(langinfo YESEXPR);
my $yesexpr = langinfo(YESEXPR());
for (;;) {
systemcall('/usr/bin/chfn', $new_name);
# Translators: [y/N] has to be replaced by values defined in your
# locale. You can see by running locale yesexpr which regular
# expression will be checked to find positive answer.
print (gtx(Is the information correct? [y/N] ));
chop ($answer=STDIN);
last if ($answer =~ m/$yesexpr/o);
}
Of course this comment may be improved ;)
You need to call xgettext with the -c flag so that comments are inserted
into PO files.
Other minor i18n problems with your package:
* the gtx() function is useless and confuses xgettext
* Some msgid are concatenated to build a full sentence. This is bad.
and in this case only the first part was present in PO files.
Here is a patch. I tested it, but of course it may introduce bugs,
please review it carefully ;)
Denis
diff -ur adduser-3.74.orig/AdduserCommon.pm adduser-3.74/AdduserCommon.pm
--- adduser-3.74.orig/AdduserCommon.pm 2005-10-18 16:20:00.0 +0200
+++ adduser-3.74/AdduserCommon.pm 2005-10-21 23:49:30.0 +0200
@@ -10,7 +10,7 @@
# Ian A. Murdock [EMAIL PROTECTED]
#
[EMAIL PROTECTED] = qw(invalidate_nscd gtx dief warnf read_config
get_users_groups get_group_members s_print s_printf systemcall);
[EMAIL PROTECTED] = qw(invalidate_nscd dief warnf read_config get_users_groups
get_group_members s_print s_printf systemcall);
sub invalidate_nscd {
# Check if we need to do make -C /var/yp for NIS
@@ -56,10 +56,6 @@
}
}
-sub gtx {
-return gettext( join , @_);
-}
-
sub dief {
my ($form,@argu)[EMAIL PROTECTED];
printf STDERR $0: $form,@argu;
@@ -80,7 +76,7 @@
my ($var, $lcvar, $val);
if (! -f $conf_file) {
- printf gtx(%s: `%s' doesn't exist. Using defaults.\n),$0,$conf_file
if $verbose;
+ printf gettext(%s: `%s' doesn't exist. Using
defaults.\n),$0,$conf_file if $verbose;
return;
}
@@ -90,12 +86,12 @@
next if /^#/ || /^\s*$/;
if ((($var, $val) = /^\s*(\S+)\s*=\s*(.*)/) != 2) {
- warnf gtx(Couldn't parse `%s':%s.\n),$conf_file,$.;
+ warnf gettext(Couldn't parse `%s':%s.\n),$conf_file,$.;
next;
}
$lcvar = lc $var;
if (!defined($configref-{$lcvar})) {
- warnf gtx(Unknown variable `%s' at `%s':%s.\n),$var,$conf_file,$.;
+ warnf gettext(Unknown variable `%s' at
`%s':%s.\n),$var,$conf_file,$.;
next;
}
diff -ur adduser-3.74.orig/adduser adduser-3.74/adduser
--- adduser-3.74.orig/adduser 2005-10-19 15:09:09.0 +0200
+++ adduser-3.74/adduser2005-10-21 23:45:02.0 +0200
@@ -48,10 +48,19 @@
if ($@) {
*setlocale = sub { return 1 };
}
+eval {
+ require I18N::Langinfo;
+ import I18N::Langinfo qw(langinfo YESEXPR);
+};
+if ($@) {
+ *langinfo = sub { ^[yY] };
+ *YESEXPR = sub { 1 };
+}
}
setlocale(LC_MESSAGES, );
textdomain(adduser);
+my $yesexpr = langinfo(YESEXPR());
our $verbose = 1; # should we be verbose?
my $allow_badname = 0; # should we allow bad names?
@@ -86,7 +95,7 @@
debug = \$debugging );
# everyone can issue --help and --version, but only root can go on
-die ($0: ,gtx(Only root may add a user or group to the system.\n)) if ($
!= 0);
+die ($0: ,gettext(Only root may add a user or group to the system.\n)) if
($ != 0);
if( defined($configfile) ) { @defaults = ($configfile); }
@@ -106,30 +115,30 @@
while (defined($arg = shift(@ARGV))) {
if (defined($names[0]) $arg =~ /^--/) {
