Re: Serious internationalization problems: revert?

Re: Serious internationalization problems: revert?

[Richard Kenner]

    What target is this 

alphaev56-dec-osf4.0c

    and does it have a system-provided libiconv?

I can't find it, but I might not be looking in the right places.

Re: Serious internationalization problems: revert?

[gcc]

On Fri, Oct 19, 2001 at 01:26:56PM -0400, Richard Kenner wrote:
>     What target is this 
> 
> alphaev56-dec-osf4.0c
> 
>     and does it have a system-provided libiconv?
> 
> I can't find it, but I might not be looking in the right places.

Look in /usr/shlib. 4.0D has /usr/shlib/libiconv.so.

-- 
albert chin (china@thewrittenword.com)

Re: Serious internationalization problems: revert?

[Zack Weinberg]

On Fri, Oct 19, 2001 at 11:36:12AM -0400, Richard Kenner wrote:
>     > (2) Some configurations complain that make can't find -liconv, since
>     >     it gets erroneously considered a dependency.
> 
>     I need more information in order to understand where this is coming
>     from.  It is supposed to notice the absence of libiconv and not use it.
> 
> We have:
> 
> INTLLIBS = $(top_builddir)/intl/libintl.a -liconv
> 
> But then LIBDEPS have $(INTLLIBS), so make tries to find a rule to make
> "-liconv".

Ah, I understand.  What target is this and does it have a
system-provided libiconv?

zw

Re: Serious internationalization problems: revert?

[Richard Kenner]

    You mean libintl.h, right? I just checked in a patch which should
    resolve this.

Yes.  Good.

    > (2) Some configurations complain that make can't find -liconv, since
    >     it gets erroneously considered a dependency.

    I need more information in order to understand where this is coming
    from.  It is supposed to notice the absence of libiconv and not use it.

We have:

INTLLIBS = $(top_builddir)/intl/libintl.a -liconv

But then LIBDEPS have $(INTLLIBS), so make tries to find a rule to make
"-liconv".

Re: Serious internationalization problems: revert?

[rbrown64]

>A recent change introduced three problems:

>(1) In some configurations, intl.h is not found.
>(2) Some configurations complain that make can't find -liconv, since
    it gets erroneously considered a dependency.
>(3) There are dozens of warnings about missing declarations and format
    arg mismatches.

>Is anybody working on this or is it time to consider reverting those
changes?

Wrt (1) at least CVS Wed Oct 17 23:03:18 UTC 2001 is working for me
on i686-unknown-sysv5UnixWare7.1.0 which had been failing for me before.


I needed to follow Zack's instructions, running autoconf and autoheader
again
and also found I needed to remove config.cache. make distclean didn't cut
it.
Maybe maintainerclean would have.

My setup is screwy though. Hope this helps

--- 1011/warning.log    Fri Oct 12 12:02:09 2001 (UnixWare 7.1 3.1)
+++ ./warning.log       Sat Oct 20 01:23:27 2001
@@ -1,38 +1,50 @@
 Counting all warnings,
-there are 89 warnings in stage3 of this bootstrap.
+there are 139 warnings in stage3 of this bootstrap.

 Number of warnings per file:
+28 gcc/intl/gettextP.h
 23 gcc/combine.c
+18 gcc/intl/loadmsgcat.c
 18 gcc/dwarfout.c
-8 gcc/langhooks.c
 7 gcc/varasm.c
 6 gcc/cp/pt.c
+5 gcc/intl/localealias.c
 5 gcc/cse.c
+4 gcc/intl/l10nflist.c
 4 gcc/gcc.c
 2 gcc/tlink.c
 2 gcc/objc/objc-act.c
-2 gcc/java/lang.c
+2 gcc/intl/localcharset.c
+2 gcc/intl/dcigettext.c
 2 gcc/fold-const.c
-2 gcc/f/com.c
+1 plural.y
+1 insn-emit.c
 1 gcc/rtlanal.c
+1 gcc/regmove.c
 1 gcc/objc/lang-specs.h
 1 gcc/java/jvspec.c
-1 gcc/intl/localealias.c
-1 gcc/intl/dcgettext.c
 1 gcc/f/lang-specs.h
+1 gcc/dwarf2out.c
 1 gcc/dwarf2asm.c
+1 gcc/cp/lex.c
 1 gcc/cp/lang-specs.h

 Number of warning types:
-54 comparison between signed and unsigned
-8 function declaration isn't a prototype
+57 comparison between signed and unsigned
+17 function declaration isn't a prototype
+15 ISO C forbids zero-size array `???'
+12 ISO C89 forbids specifying subobject to initialize
+9 subscript has type `char'
 7 string length `???' is greater than the length `???' ISO C89 compilers
are re
quired to support
 6 ??? format, ??? arg (arg ???)
-3 missing initializer
-3 (near initialization for `???')
+2 traditional C rejects initialization of unions
 2 passing arg ??? of `???' discards qualifiers from pointer target type
+2 missing initializer
 2 implicit declaration of function `???'
+2 (near initialization for `???')
+1 unused parameter `???'
 1 suggest not using #elif in traditional C
 1 signed and unsigned type in conditional expression
+1 right shift count >= width of type
 1 pointer targets in passing arg ??? of `???' differ in signedness
 1 `???' defined but not used

Re: Serious internationalization problems: revert?

[Zack Weinberg]

On Fri, Oct 19, 2001 at 07:27:27AM -0400, Richard Kenner wrote:
> A recent change introduced three problems:
> 
> (1) In some configurations, intl.h is not found.

You mean libintl.h, right? I just checked in a patch which should
resolve this.

> (2) Some configurations complain that make can't find -liconv, since
>     it gets erroneously considered a dependency.

I need more information in order to understand where this is coming
from.  It is supposed to notice the absence of libiconv and not use
it.

> (3) There are dozens of warnings about missing declarations and format
>     arg mismatches.

This is because upstream gettext is coded to a weaker set of warning
switches than GCC uses.  I'll be looking at this after the critical
problems are resolved.

zw

Re: Serious internationalization problems: revert?

[Richard Kenner]

    (Neither your nor my reports concerning this have received any responses,
    have they?)

Mine have not.

Re: Serious internationalization problems: revert?

[Richard Earnshaw]

> On Fri, 19 Oct 2001, Richard Kenner wrote:
> > Is anybody working on this or is it time to consider reverting those
> > changes?
> 
> The latter, as far as I am concerned.
> 
> (Neither your nor my reports concerning this have received any responses,
> have they?)
> 
> Gerald
> 

I believe Zack was working on it.

http://gcc.gnu.org/ml/gcc-patches/2001-10/msg00619.html

I've not tried it to see if that fixed the problem, and I don't know if 
the patch has been committed yet.

R.

Re: Serious internationalization problems: revert?

[Gerald Pfeifer]

On Fri, 19 Oct 2001, Richard Kenner wrote:
> Is anybody working on this or is it time to consider reverting those
> changes?

The latter, as far as I am concerned.

(Neither your nor my reports concerning this have received any responses,
have they?)

Gerald

Serious internationalization problems: revert?

[Richard Kenner]

A recent change introduced three problems:

(1) In some configurations, intl.h is not found.
(2) Some configurations complain that make can't find -liconv, since
    it gets erroneously considered a dependency.
(3) There are dozens of warnings about missing declarations and format
    arg mismatches.

Is anybody working on this or is it time to consider reverting those changes?