sys/mount.h not C++ clean on kfreebsd

2012-02-25 Thread peter green
Libreoffice failed to build on kfreebsd-i386 and kfreebsd-amd64 with the 
following error.


Compiling: sal/osl/unx/file_volume.cxx
In file included from 
/build/buildd-libreoffice_3.4.5-3-kfreebsd-amd64-ukEv52/libreoffice-3.4.5/libreoffice-build/build/libreoffice-3.4.5.2/sal/osl/unx/file_volume.cxx:77:0:
/usr/include/x86_64-kfreebsd-gnu/sys/mount.h: In function 'int 
getvfsbyname(const char*, xvfsconf*)':
/usr/include/x86_64-kfreebsd-gnu/sys/mount.h:398:23: error: invalid conversion 
from 'void*' to 'xvfsconf*' [-fpermissive]
/build/buildd-libreoffice_3.4.5-3-kfreebsd-amd64-ukEv52/libreoffice-3.4.5/libreoffice-build/build/libreoffice-3.4.5.2/sal/osl/unx/file_volume.cxx:
 At global scope:
/build/buildd-libreoffice_3.4.5-3-kfreebsd-amd64-ukEv52/libreoffice-3.4.5/libreoffice-build/build/libreoffice-3.4.5.2/sal/osl/unx/file_volume.cxx:1153:17:
 warning: unused parameter 'pszPath' [-Wunused-parameter]
/build/buildd-libreoffice_3.4.5-3-kfreebsd-amd64-ukEv52/libreoffice-3.4.5/libreoffice-build/build/libreoffice-3.4.5.2/sal/osl/unx/file_volume.cxx:1153:17:
 warning: unused parameter 'pItem' [-Wunused-parameter]
/build/buildd-libreoffice_3.4.5-3-kfreebsd-amd64-ukEv52/libreoffice-3.4.5/libreoffice-build/build/libreoffice-3.4.5.2/sal/osl/unx/file_volume.cxx:1160:17:
 warning: unused parameter 'pDevice' [-Wunused-parameter]
/build/buildd-libreoffice_3.4.5-3-kfreebsd-amd64-ukEv52/libreoffice-3.4.5/libreoffice-build/build/libreoffice-3.4.5.2/sal/osl/unx/file_volume.cxx:555:35:
 warning: 'oslVolumeDeviceHandleImpl* osl_newVolumeDeviceHandleImpl()' defined 
but not used [-Wunused-function]
/build/buildd-libreoffice_3.4.5-3-kfreebsd-amd64-ukEv52/libreoffice-3.4.5/libreoffice-build/build/libreoffice-3.4.5.2/sal/osl/unx/file_volume.cxx:579:13:
 warning: 'void osl_freeVolumeDeviceHandleImpl(oslVolumeDeviceHandleImpl*)' 
defined but not used [-Wunused-function]
/build/buildd-libreoffice_3.4.5-3-kfreebsd-amd64-ukEv52/libreoffice-3.4.5/libreoffice-build/build/libreoffice-3.4.5.2/sal/osl/unx/file_volume.cxx:1153:17:
 warning: 'sal_Bool osl_getFloppyMountEntry(const sal_Char*, 
oslVolumeDeviceHandleImpl*)' defined but not used [-Wunused-function]
/build/buildd-libreoffice_3.4.5-3-kfreebsd-amd64-ukEv52/libreoffice-3.4.5/libreoffice-build/build/libreoffice-3.4.5.2/sal/osl/unx/file_volume.cxx:1160:17:
 warning: 'sal_Bool osl_isFloppyMounted(oslVolumeDeviceHandleImpl*)' defined 
but not used [-Wunused-function]
dmake:  Error code 1, while making '../../unxkfgx6.pro/obj/file_volume.obj'
Retrying 
/build/buildd-libreoffice_3.4.5-3-kfreebsd-amd64-ukEv52/libreoffice-3.4.5/libreoffice-build/build/libreoffice-3.4.5.2/sal/osl/unx


A simple test program including the header (and with an empty main 
function) compiles fine when built as C but fails when built as C++. So 
it looks like this is a case of the header not being C++ clean (implicit 
conversion from void* to other pointer types is allowed in C but not in 
C++). Presumablly the fix is to add an explicit typecast.


I would file a bug report but packages.debian.org doesn't seem to want 
to tell me what package sys/mount.h is in.



--
To UNSUBSCRIBE, email to debian-openoffice-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/4f48d24d.60...@p10link.net



Re: sys/mount.h not C++ clean on kfreebsd

2012-02-25 Thread Rene Engelhard
Hi,

On Sat, Feb 25, 2012 at 12:21:33PM +, peter green wrote:
 Compiling: sal/osl/unx/file_volume.cxx
 In file included from 
 /build/buildd-libreoffice_3.4.5-3-kfreebsd-amd64-ukEv52/libreoffice-3.4.5/libreoffice-build/build/libreoffice-3.4.5.2/sal/osl/unx/file_volume.cxx:77:0:
 /usr/include/x86_64-kfreebsd-gnu/sys/mount.h: In function 'int 
 getvfsbyname(const char*, xvfsconf*)':
 /usr/include/x86_64-kfreebsd-gnu/sys/mount.h:398:23: error: invalid 
 conversion from 'void*' to 'xvfsconf*' [-fpermissive]

Already pointed out on 
http://lists.debian.org/debian-bsd/2012/02/msg00061.html
see also
http://lists.debian.org/debian-bsd/2012/02/msg00080.html

And aurel32 promised that upload for this weekend :)

 I would file a bug report but packages.debian.org doesn't seem to
 want to tell me what package sys/mount.h is in.

libc0.1-dev

Regards,

Rene


-- 
To UNSUBSCRIBE, email to debian-openoffice-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/20120225125603.gf17...@rene-engelhard.de



Re: libreoffice, mingw-w64, gcc-mingw-w64 and gnat-4.6 on armhf

2012-02-25 Thread Luke Kenneth Casson Leighton
On Fri, Feb 3, 2012 at 5:48 AM, Stephen Kitt st...@sk2.org wrote:
 Hi Peter,

 On Fri, Feb 03, 2012 at 02:36:17AM +, peter green wrote:
 Libreoffice hasn't yet been built on armhf. I consider libreoffice
 to be a reasonablly important package and one that we need to get in
 before we can claim we have a reasonablly complete port.

 [...]

 This (build-)dependency chain leaves me with a few questions
 1: what is the current status of gnat-4.6 on armhf? does an upload
 look likely any time soon?
 2: why does libreoffice need mingw-w64 in the first place?

 libreoffice uses mingw-w64 to build a DLL, unowinreg.dll, which is
 provided in the libreoffice-dev package. As I understand it the DLL
 itself isn't used on Debian, but it is provided by the SDK because it
 is supposed to be bundled with plugins which need to access the
 registry, and therefore to be able to correctly build shippable
 plugins using Debian the SDK packages need to provide the DLL.

 it's even more hilarious than that: it's actually because java can't
access windows registry functions, so someone wrote a c-based DLL
which java *can* bind to.  the fact that the end-result of the
mingw-w64 cross-compiler output would be an x86 64-bit DLL, which
simply wouldn't even run on an ARM processor anyway seems to have
entirely escaped everyone's attention.

 i describe the chain here, and have made a request on behalf of the
sanity of the debian-arm team that the libreoffice developers consider
adding a compile-time switch to remove the complete mental brain-fart
retardation from the software for which they are responsible:

 https://bugs.freedesktop.org/show_bug.cgi?id=46614

let's see if they have a sense of humour, eh?

l.


-- 
To UNSUBSCRIBE, email to debian-openoffice-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: 
http://lists.debian.org/CAPweEDyj=rqmiekmqe2z1u03hvr8_27dkdsjzfcyjtma2jl...@mail.gmail.com



Re: libreoffice, mingw-w64, gcc-mingw-w64 and gnat-4.6 on armhf

2012-02-25 Thread Rene Engelhard
Hi,

On Sat, Feb 25, 2012 at 01:49:02PM +, Luke Kenneth Casson Leighton wrote:
  it's even more hilarious than that: it's actually because java can't
 access windows registry functions, so someone wrote a c-based DLL
 which java *can* bind to.  the fact that the end-result of the

Yes, that is the point.

 mingw-w64 cross-compiler output would be an x86 64-bit DLL, which

No, it's a 32 bit dll.

# file /usr/share/libreoffice/sdk/classes/win/unowinreg.dll
/usr/share/libreoffice/sdk/classes/win/unowinreg.dll: PE32 executable (DLL) 
(console) Intel 80386 (stripped to external PDB), for MS Windows

i686-w64-mingw32-g++ is called.

 simply wouldn't even run on an ARM processor anyway seems to have
 entirely escaped everyone's attention.

No. In contast, Stephen said it correctly.

[...] isn't used on Debian, but it is provided by the SDK because it is 
supposed to be
bundled with plugins [...] and therefore to be able to correctly build 
shippable
plugins using Debian the SDK packages need to provide the DLL.

  i describe the chain here, and have made a request on behalf of the
 sanity of the debian-arm team that the libreoffice developers consider
 adding a compile-time switch to remove the complete mental brain-fart
 retardation from the software for which they are responsible:
 
  https://bugs.freedesktop.org/show_bug.cgi?id=46614

To be fair, that all is inherited from OOo.
And it's a non-issue now that we *do* have mingw-w64 on armhf.

 let's see if they have a sense of humour, eh?

If the bug is formulated like the nonsense in this post I won't believe so.

Regards,

Rene


-- 
To UNSUBSCRIBE, email to debian-openoffice-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/20120225140424.gg17...@rene-engelhard.de



Re: libreoffice, mingw-w64, gcc-mingw-w64 and gnat-4.6 on armhf

2012-02-25 Thread Luke Kenneth Casson Leighton
On Sat, Feb 25, 2012 at 2:04 PM, Rene Engelhard r...@debian.org wrote:
 Hi,

 On Sat, Feb 25, 2012 at 01:49:02PM +, Luke Kenneth Casson Leighton wrote:
  it's even more hilarious than that: it's actually because java can't
 access windows registry functions, so someone wrote a c-based DLL
 which java *can* bind to.  the fact that the end-result of the

 Yes, that is the point.

 mingw-w64 cross-compiler output would be an x86 64-bit DLL, which

 No, it's a 32 bit dll.

 yes - steven kindly privately outlined a bit more detail about what's
involved: he said that it's actually part of the developer kit.


 # file /usr/share/libreoffice/sdk/classes/win/unowinreg.dll
 /usr/share/libreoffice/sdk/classes/win/unowinreg.dll: PE32 executable (DLL) 
 (console) Intel 80386 (stripped to external PDB), for MS Windows

 i686-w64-mingw32-g++ is called.

 that's different from mingw-w64, then.

 simply wouldn't even run on an ARM processor anyway seems to have
 entirely escaped everyone's attention.

 No. In contast, Stephen said it correctly.

 actually, he didn't: in the public post he didn't mention that it was
purely for shipping with the *windows* version of libreoffice, so that
people who perform development on gnu/linux of libreoffice
applications can ship the libreoffice application with that DLL *such
that* the *windows* version of libreoffice will actually work and have
access to the windows dll.


 [...] isn't used on Debian, but it is provided by the SDK because it is 
 supposed to be
 bundled with plugins [...] and therefore to be able to correctly build 
 shippable
 plugins using Debian the SDK packages need to provide the DLL.

  i describe the chain here, and have made a request on behalf of the
 sanity of the debian-arm team that the libreoffice developers consider
 adding a compile-time switch to remove the complete mental brain-fart
 retardation from the software for which they are responsible:

  https://bugs.freedesktop.org/show_bug.cgi?id=46614

 To be fair, that all is inherited from OOo.
 And it's a non-issue now that we *do* have mingw-w64 on armhf.

 hurrah!

 let's see if they have a sense of humour, eh?

 If the bug is formulated like the nonsense in this post I won't believe so.

 your bullying and lack of forgiveness is duly noted.

 l.


--
To UNSUBSCRIBE, email to debian-openoffice-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: 
http://lists.debian.org/CAPweEDyN1EbYuENW8Ph=u5xayzmq8_d2mchwrlpmrp7ezkp...@mail.gmail.com



Re: sys/mount.h not C++ clean on kfreebsd

2012-02-25 Thread Robert Millan
El 25 de febrer de 2012 12:56, Rene Engelhard r...@debian.org ha escrit:
 I would file a bug report but packages.debian.org doesn't seem to
 want to tell me what package sys/mount.h is in.

 libc0.1-dev

There's already #660397 and #660401.

-- 
Robert Millan


-- 
To UNSUBSCRIBE, email to debian-openoffice-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: 
http://lists.debian.org/caofdtxoawacsese6fkeec87kk-jelepq2sow11ayzl_--xd...@mail.gmail.com



Re: libreoffice, mingw-w64, gcc-mingw-w64 and gnat-4.6 on armhf

2012-02-25 Thread peter green

Luke Kenneth Casson Leighton wrote:

i686-w64-mingw32-g++ is called.



 that's different from mingw-w64, then.
  

No it's *part of* mingw-w64.

Mnigw-w64 is a fork of mingw32 and provides toolchains targetting both 
32-bit and 64-bit windows. These toolchains (at least in debian, not 
sure about upstream) have triplets i686-w64-mingw32 and x86_64-w64-mingw32 .



To be fair, that all is inherited from OOo.
And it's a non-issue now that we *do* have mingw-w64 on armhf.



 hurrah!
  
Yeah,  I provided the gcc-mingw-w64 maintainers with a patch that 
disabled gnat support on armhf allowing the rest of the package to be 
built and the libreoffice build to be tried.


Thanks to everyone who helped get libreoffice into armhf unstable.



--
To UNSUBSCRIBE, email to debian-openoffice-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/4f49688e.6040...@p10link.net



Bug#661294: libreoffice FTBFS on alpha because of unsatisfiable build-depends on libgraphite2-dev

2012-02-25 Thread Michael Cree
Source: libreoffice
Version: 1:3.4.5-3
Severity: normal
User: debian-al...@lists.debian.org
Usertags: alpha
X-Debbugs-CC: debian-al...@lists.debian.org

libreoffice can't be built on the Alpha architecture because graphite2
FTBFS, thus the build-dependency on libgraphite2-dev cannot be
satisfied.  I see that armel and sparc have the same problem and that it
has been resolved by removing the build-dependency on libgraphite2-dev
for those architectures.  I am asking that the same be done for Alpha.

Please note that openoffice has built in the past on Alpha---about nine
months ago we had an installable and also working version of openoffice.

Thanks
Michael



-- 
To UNSUBSCRIBE, email to debian-openoffice-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/4f4983ab.10...@orcon.net.nz