Re: Gettext version bump [was Re: make distcheck warnings]

2005-11-13 Thread Pavel Roskin
Quoting Leonard den Ottolander [EMAIL PROTECTED]:

 Hi Pavel,

 On Sat, 2005-11-12 at 19:51 -0500, Pavel Roskin wrote:
  What kind of breakage do you see or expect to see?

 Being unable to build mc on such legacy systems as the build
 requirements are not met. This version bump also bumps the version
 requirements for automake and autoconf.

You are supposed to build from a tarball on such systems, not from CVS.

  Darwin support without that ugly kludge in configure.ac.  Less warnings
 (there
  are still plenty of them in the gettext code if you run my mctest, but
 neither
  of them seems to indicate a real problem).

 I'd rather live with a few redundant warnings than without a more recent
 mc on such legacy systems.

I'm yet to see any complaints from anyone who actually tried that.

If you insist on building from CVS on legacy systems, I could write a wrapper or
replacement for autopoint to ignore the version requirements for gettext.  I
could also ask gettext maintainers to allow support for the at least operator
for gettext version - I think it should be safe for mc.

--
Regards,
Pavel Roskin
___
Mc-devel mailing list
http://mail.gnome.org/mailman/listinfo/mc-devel


Re: Gettext version bump [was Re: make distcheck warnings]

2005-11-13 Thread Leonard den Ottolander
Hi Pavel,

On Sun, 2005-11-13 at 07:44 -0500, Pavel Roskin wrote:
 You are supposed to build from a tarball on such systems, not from CVS.

I wasn't aware the dependency on the particular gettext version is only
there for the pre distchecked code. But indeed a build from tarball on a
RHL 9 seems not to be affected by the gettext version. (It bombs out on
an error in editcmd.c, but that is an issue with the UTF8 patch which
I'll take up downstream.)

Thanks for clearing this up.

Leonard.

-- 
mount -t life -o ro /dev/dna /genetic/research


___
Mc-devel mailing list
http://mail.gnome.org/mailman/listinfo/mc-devel


Re: make distcheck warnings

2005-11-12 Thread Leonard den Ottolander
Hello Pavel,

On Thu, 2005-11-10 at 20:55 -0500, Pavel Roskin wrote:
 I chose version 0.14.3 was chosen because Fedora Core 4 supplies it.  It
 should be old enough for the core developers to have it, and new enough
 to have most bugs fixed.

I'm not so sure it's such a good idea to have the build depend on such a
recent version of gettext. You might fix a few bugs, but you break
compatibility with a lot of legacy systems. What exactly is it we need
version 0.14.3 for?

Leonard.

-- 
mount -t life -o ro /dev/dna /genetic/research


___
Mc-devel mailing list
http://mail.gnome.org/mailman/listinfo/mc-devel


Re: make distcheck warnings

2005-11-12 Thread Pavel Roskin
Quoting Leonard den Ottolander [EMAIL PROTECTED]:

 Hello Pavel,

 On Thu, 2005-11-10 at 20:55 -0500, Pavel Roskin wrote:
  I chose version 0.14.3 was chosen because Fedora Core 4 supplies it.  It
  should be old enough for the core developers to have it, and new enough
  to have most bugs fixed.

 I'm not so sure it's such a good idea to have the build depend on such a
 recent version of gettext. You might fix a few bugs, but you break
 compatibility with a lot of legacy systems.

What kind of breakage do you see or expect to see?  Could you please provide
more details?  I actually think newer versions of gettext should work better
with legacy systems, since the fixes are incorporated into gettext runtime over
time.

 What exactly is it we need
 version 0.14.3 for?

Darwin support without that ugly kludge in configure.ac.  Less warnings (there
are still plenty of them in the gettext code if you run my mctest, but neither
of them seems to indicate a real problem).

Actually, the CVS version of gettext supports message contexts, so it would be
nice to switch to it soon after it's released.

--
Regards,
Pavel Roskin
___
Mc-devel mailing list
http://mail.gnome.org/mailman/listinfo/mc-devel


make distcheck warnings

2005-11-10 Thread Leonard den Ottolander
Hi,

make distcheck reports the following warnings:

msgmerge: ta.po: warning: Charset TSCII is not a portable encoding
name.
/usr/bin/msgfmt: ta.po: warning: Charset TSCII is not a portable
encoding name.
configure: WARNING: Rejecting S-Lang with UTF-8 support, it's not fully
supported yet
../../intl/intl-compat.c:24:1: warning: _INTL_REDIRECT_MACROS
redefined
../config.h:644:1: warning: this is the location of the previous
definition
../config.h:641:1: warning: _GNU_SOURCE redefined
../../intl/loadmsgcat.c:23:1: warning: this is the location of the
previous definition
../config.h:641:1: warning: _GNU_SOURCE redefined
../../intl/localealias.c:23:1: warning: this is the location of the
previous definition
../config.h:641:1: warning: _GNU_SOURCE redefined
../../intl/l10nflist.c:23:1: warning: this is the location of the
previous definition
../config.h:641:1: warning: _GNU_SOURCE redefined
../../intl/dcigettext.c:23:1: warning: this is the location of the
previous definition
msgmerge: ta.po: warning: Charset TSCII is not a portable encoding
name.

Leonard.
-- 
mount -t life -o ro /dev/dna /genetic/research


___
Mc-devel mailing list
http://mail.gnome.org/mailman/listinfo/mc-devel


Re: make distcheck warnings

2005-11-10 Thread Pavel Roskin
Hello, Leonard!

On Thu, 2005-11-10 at 22:42 +0100, Leonard den Ottolander wrote: 
 Hi,
 
 make distcheck reports the following warnings:
 
 msgmerge: ta.po: warning: Charset TSCII is not a portable encoding
 name.
 /usr/bin/msgfmt: ta.po: warning: Charset TSCII is not a portable
 encoding name.

We cannot fix it in mc.  It was discussed many times.  Google for mc
TSCII.

TSCII is not standard, but it's actually used.  Either TSCII should be
standardized, or Unicode should be used.  If I remember correctly, there
are some problems with Tamil support in Unicode.  Maybe the consoles
fonts is not so good.  I don't think we should pressure Tamil users to
follow standards if they don't have good Unicode fonts.

 configure: WARNING: Rejecting S-Lang with UTF-8 support, it's not fully
 supported yet
 ../../intl/intl-compat.c:24:1: warning: _INTL_REDIRECT_MACROS
 redefined
 ../config.h:644:1: warning: this is the location of the previous
 definition

The redefinition of _INTL_REDIRECT_MACROS is fixed in 0.14.3 (it might
have been fixed much earlier).

The problem with redirection was reported on Darwin, where
__USER_LABEL_PREFIX__ wasn't defined.  The problem was fixed in gettext
immediately after 0.11.5.  Since mc supported older versions of gettext,
a workaround was provided.

I also noticed that the test used in mc was incorrect - it used nested
functions instead of global functions used by gettext.  I fixed that
initially.

But then I settled for a better approach - mc is using gettext 0.14.3
now, and MC_ASM_LABELS is gone.

I chose version 0.14.3 was chosen because Fedora Core 4 supplies it.  It
should be old enough for the core developers to have it, and new enough
to have most bugs fixed.

 ../config.h:641:1: warning: _GNU_SOURCE redefined
 ../../intl/loadmsgcat.c:23:1: warning: this is the location of the
 previous definition

This is fixed by using AC_GNU_SOURCE.  Since it appeared in Autoconf
2.54, configure.ac requires it now.

-- 
Regards,
Pavel Roskin

___
Mc-devel mailing list
http://mail.gnome.org/mailman/listinfo/mc-devel