An Octave user reported the following link error when building Octave
on a Windows system with MinGW:
libtool: compile: g++ -shared-libgcc -DHAVE_CONFIG_H -I.
-I../../../hg/octave-work/liboctave -I..
-Ic:/Programs/OctaveBuild/include -Ic:/Programs/WinDevTools/include
-Ic:/Programs/GnuWin32/
Looking at lib/glob.c in the gnulib sources, there is some
Windows-specific code, so it looks like it is intended to work on
Windows systems, but there are some things that don't look quite
right. For example, it does not seem to uniformly use backslash as a
directory separator, and if WINDOWS32 i
I'm using the glob module in GNU Octave and a Windows user complained
that linking failed with an undefined reference to getlogin.
The problem appears to be that getlogin_r is called from glob, and the
gnulib replacement for getlogin_r calls getlogin unconditionally, but
getlogin doesn't exist on
According to Eric Blake on 1/6/2010 2:00 PM:
> Jim Meyering meyering.net> writes:
>> Then I realized that the patch above would probably be enough.
>> The GFDL version number will not be changing often.
>
> Fair enough. So, do we apply your patch, or should I try testing/tweaking it
> further?
According to Ian Beckwith on 1/6/2010 6:05 PM:
> Hi,
>
> select (LGPLv2+) and nanosleep (GPL) both need a dependency on
> sockets (LGPLv2+).
nanosleep already depends on select. And:
./gnulib-tool --with-tests --test select
already pulls in sockets as one of its indirect dependencies.
Oh, I s
According to Ian Beckwith on 1/6/2010 6:05 PM:
>> during build of coreutils-8.1:
>>
>> at-func2.c: In function 'at_func2':
>> at-func2.c:113: warning: attempt to free a non-heap object 'proc_buf1'
>>
>> IMO, the right fix is
>>
>> --- lib/at-func2.c
>> +++ lib/at-func2.c
>> @@ -110,7 +110,7 @@
>>
Hi,
select (LGPLv2+) and nanosleep (GPL) both need a dependency on
sockets (LGPLv2+).
Reported against gnuit:
> From: "Nelson H. F. Beebe"
> Several months ago, I successfully built gnuit-4.9.5 on 20+ flavors of
> Unix. Today, I tried a recent addition to our test laboratory: MirOS
> (http://
Copyright (C) 1992, 1993, 1994, 2004, 2005, 2006, 2007, 2008, 2009, 2010
Free Software Foundation, Inc.
[becoming]
Copyright (C) 2001-2010 Free Software Foundation, Inc.
Such a change would be quite undesirable, IMHO.
So I'm glad you found another approach.
Hi Eric,
* Eric Blake wrote on Wed, Jan 06, 2010 at 10:15:43PM CET:
> Ralf Wildenhues gmx.de> writes:
>
> > > Do we need to be using AC_USE_SYSTEM_EXTENSIONS for pread?
> >
> > I don't think AC_USE_SYSTEM_EXTENSIONS defines _XOPEN_SOURCE. I'm not
> > sure it should though, it also disables som
Ralf Wildenhues gmx.de> writes:
> > Do we need to be using AC_USE_SYSTEM_EXTENSIONS for pread?
>
> I don't think AC_USE_SYSTEM_EXTENSIONS defines _XOPEN_SOURCE. I'm not
> sure it should though, it also disables some APIs on some systems.
On Linux, AC_USE_SYSTEM_EXTENSIONS defines _GNU_SOURCE,
Ralf Wildenhues gmx.de> writes:
>
> The fts-lgpl module fails to build:
>
> I can make the test finish compilation by copying over
> fcntl--.h fcntl-safer.h dirent--.h dirent-safer.h
>
> but those files come from GPL'ed modules, and it doesn't seem like the
> right fix anyway.
Correct. The
Jim Meyering meyering.net> writes:
> > But I'm not fluent enough with sed to
> > quickly write a script that can sanitize both styles into a single
recognizable
> > line:
> >
>
> This should do it
>
> perl -0777 -pe 's/^Copyright.+?Free\sSoftware\sFoundation,\sInc\.//ms'
Indeed - once we re
Ralf Wildenhues gmx.de> writes:
>
> The xlist and the xoset modules fail to gnulib --test due to missing
> headers; the patch below fixes that. OK to commit?
>
> Thanks,
> Ralf
>
> Include missing headers for xlist and xoset modules.
> * modules/xlist (Files): Add lib/gl_list.h.
The fts-lgpl module fails to build:
$ gnulib-tool --with-tests --test fts-lgpl
[...]
make[3]: Entering directory `/tmp/gnulib/build/fts-lgpl/gllib'
gcc -DHAVE_CONFIG_H -I. -I/tmp/gnulib/build/../testdir/fts-lgpl/gllib -I..
-g -O2 -MT fts.o -MD -MP -MF .deps/fts.Tpo -c -o fts.o
/tmp/gnulib/bu
Eric Blake wrote:
> I noticed that NEWS_hash in maint.mk doesn't work for m4. Why? Because m4
> uses long-hand copyright years, and lists enough years that the
> regex '^Copyright.*Free Software' no longer matches. In other words, adding
> 2010 to the copyright changed the old news hash.
>
> For
The xlist and the xoset modules fail to gnulib --test due to missing
headers; the patch below fixes that. OK to commit?
Thanks,
Ralf
Include missing headers for xlist and xoset modules.
* modules/xlist (Files): Add lib/gl_list.h.
* modules/xoset (Files): Add lib/gl_oset.h
Hi Eric,
* Eric Blake wrote on Wed, Jan 06, 2010 at 03:12:34PM CET:
> According to Ralf Wildenhues on 1/5/2010 11:44 PM:
> > The pread failure is without a patch; it requires
> > #define _XOPEN_SOURCE 500
> >
> > here but I'm not sure what the right gnuliby fix is: an AC_CHECK_DECL
> > somewher
I noticed that NEWS_hash in maint.mk doesn't work for m4. Why? Because m4
uses long-hand copyright years, and lists enough years that the
regex '^Copyright.*Free Software' no longer matches. In other words, adding
2010 to the copyright changed the old news hash.
For that matter, NEWS_hash is
According to Ralf Wildenhues on 1/5/2010 11:44 PM:
> SIGNATURE_CHECK fails a few tests due to undefined ssize_t, and pread,
> on my GNU/Linux system.
>
> Not sure if the patch is right, or if the ssize_t module is needed.
>
> The pread failure is without a patch; it requires
> #define _XOPEN_SO
Eric Blake wrote:
...
> So, how about this NEWS item?
>
> diff --git a/NEWS b/NEWS
...
> + cp -p, install -p, mv, and touch -c could trigger a spurious error
> + message when using new glibc coupled with an old kernel.
> + [bug introduced in coreutils-6.12].
Thanks for the research and summary.
According to Jim Meyering on 1/5/2010 2:07 PM:
>> --- Comment #5 from Ondrej Vasik 2010-01-05 14:29:13
>> EDT ---
>> Ah - even the latest gnulib doesn't handle this new recent change - so touch
>> from coreutils might be affected as well as mentioned in debian bugzilla from
>> comment #4. Theref
libunistring is starting to be included in many GNU/Linux distributions,
but still most programs using gnulib's unicode modules have a nontrivial
amount of code that could be eliminated if the system libunistring was
used instead.
So maybe we could do the following:
1) change libunistring to
Ralf Wildenhues wrote:
> * Jim Meyering wrote on Tue, Jan 05, 2010 at 07:09:57PM CET:
>> Paolo Bonzini wrote:
>> > On 01/04/2010 03:03 PM, Jim Meyering wrote:
>> >> However, it would be far better if AC_CHECK_FUNCS_ONCE
>> >> were to expand to e.g., ":" rather than the empty string,
>> >> so that i
Ralf Wildenhues wrote:
> Existence of leftover conftest.lnk2 file from readlink test breaks the
> gl_FUNC_SYMLINK test (wrongly returning 2 instead of 0). OK to commit?
>
> * m4/readlink.m4 (gl_FUNC_READLINK): Remove conftest.lnk2,
> to avoid failure of symlink test later.
>
> diff --g
At first, I was thinking of reverting most of that patch,
and even wrote the log and committed (locally) the result.[1]
But then I thought "What systems would suffer if I didn't?"
Only, shall we say, "challenged" systems would suffer the added
overhead. Overhead that is as yet only theoretical/
25 matches
Mail list logo
