Yaakov (Cygwin/X | 2 Jul 2009 02:55
Picon
Gravatar

Re: [1.7] KDE release plans (long)

On 23/06/2009 03:16, Ralf Habacker wrote:
> nice to hear that there are ongoing efforts :-)

I just checked into Ports SVN a snapshot of my .cygport's and patches 
for KDE 4.2.4, so now you can see what I've done:

http://cygwin-ports.svn.sourceforge.net/viewvc/cygwin-ports/ports/trunk/kde/

As for packaging, I've almost got kdelibs figured out; the rest is 
nowhere near finished.

> It is possible to give related tools a different name on the cygwin
> platform.
>
> In the kde buildsystem the variable KDE4_KCFGC_EXECUTABLE (located in
> kdelibs/cmake/modules/FindKDE4Internals.cmake) contains the name of the
> config compiler.
>
> The executable name of the config compiler could be changed in
> kdelibs\kdecore\kconfig_compiler\CMakeLists.txt by adding an OUTPUT_NAME
> property.

That much I was aware of, but it's more than just that:

1) kdelibs/CONFLICTS lists 4 more, and omitted one, which are marked 
"devel only, doesn't matter [if they collide]".

* checkXML: a simple script for invoking `meinproc{,4} --check`.  Is 
this used by either automated build system, or is this only invoked 
manually?  Are meinproc 3 & 4 compatible?
(Continue reading)

Yaakov (Cygwin/X | 2 Jul 2009 03:23
Picon
Gravatar

Re: [1.7] KDE release plans (long)

On 23/06/2009 02:29, Yaakov (Cygwin/X) wrote:
> As for a timeline, I won't dare to guess.  As I've already built almost
> everything, I can commit .cygport's and patches to SVN ASAP, but it will
> take some time to get the packaging straightened out.

Today I committed what I have for 4.2.4 to Ports SVN trunk, if anyone is 
interested.  As for the packaging and shipping, with 4.3.0 coming down 
the pipe later this month, it depends on how long it takes me to put 
together 4.2.4.

A few updates on the porting:

> * kdegraphics
> 3.5: kpovmodeler

KDE4 version available in extragear/graphics.

> * kdeutils
> 3.5: kregexpeditor

KDE4 version is in playground/utils, shipped by OpenSUSE.

> * kdewebdev
> This one is a bit tricky.  The main app of kdewebdev, quanta, won't be
> ported until 4.3 or 4.4, once kdevplatform is stable.  The other apps in
> this suite, while they can be used by themselves, are primarily KParts
> which are used in quanta.  So either I 1) just ship all 3.5 until quanta
> is ready, or possibly 2) ship kfilereplace, kimagemapeditor,
> klinkstatus, kommander, kxsldbg for 4.2, and ship quanta together with
> the KParts for 3.5.
(Continue reading)

Yaakov (Cygwin/X | 2 Jul 2009 06:37
Picon
Gravatar

Re: [1.7] KDE release plans (long)

On 01/07/2009 19:55, Yaakov (Cygwin/X) wrote:
> I just checked into Ports SVN a snapshot of my .cygport's and patches
> for KDE 4.2.4, so now you can see what I've done:
>
> http://cygwin-ports.svn.sourceforge.net/viewvc/cygwin-ports/ports/trunk/kde/
>
> As for packaging, I've almost got kdelibs figured out; the rest is
> nowhere near finished.

I have posted those patches ready for general consumption upstream:

https://bugs.kde.org/show_bug.cgi?id=130219

Ralf, could you drop in there and take a look?

Yaakov
Cygwin Ports

------------------------------------------------------------------------------
ralf.habacker | 2 Jul 2009 08:12
Picon
Favicon
Gravatar

Re: [1.7] KDE release plans (long)

Hi Yaakov,

> > I just checked into Ports SVN a snapshot of my
.cygport's and
 > patches
 > > for KDE 4.2.4, so now you can see what I've done:
 > >
 > >
 >
http://cygwin-ports.svn.sourceforge.net/viewvc/cygwin-ports/ports/trunk/kde/
 > >
 > > As for packaging, I've almost got kdelibs figured out;
the rest is
 > > nowhere near finished.
 > 
 > I have posted those patches ready for general
consumption upstream:
 > 
 > https://bugs.kde.org/show_bug.cgi?id=130219
 > 
 > Ralf, could you drop in there and take a look?
 >

patches are applied to trunk.

BTW: I have limited the last two patches to the cygwin os
because I cannot estimate if they will have side effects on
other unix platforms. Two patches need to be applied
manually because there were outdated.

(Continue reading)

Yaakov (Cygwin/X | 2 Jul 2009 09:03
Picon
Gravatar

Re: [1.7] KDE release plans (long)

On 02/07/2009 01:12, Ralf Habacker wrote:
> patches are applied to trunk.

Thanks!  Are you able to apply these to the 4.3 branch as well?

> BTW: I have limited the last two patches to the cygwin os
> because I cannot estimate if they will have side effects on
> other unix platforms.

I'm certain someone else on the bug could confirm that there would not be.

 > Two patches need to be applied manually because there were outdated.

Against trunk, very possibly.  They all did apply against 4.2.95.

Yaakov
Cygwin Ports

------------------------------------------------------------------------------
Yaakov (Cygwin/X | 3 Jul 2009 06:51
Picon
Gravatar

Re: [1.7] KDE release plans (long)

On 23/06/2009 03:16, Ralf Habacker wrote:
> The executable name of the config compiler could be changed in
> kdelibs\kdecore\kconfig_compiler\CMakeLists.txt by adding an OUTPUT_NAME
> property.

Ralf,

I have been unable to make this work.

1) I presume you meant the following:

set_target_properties(kconfig_compiler PROPERTIES OUTPUT_NAME 
kconfig_compiler4)

That creates kconfig_compiler4.exe, but it still make 
kconfig_compiler.exe.shell (no '4') which looks for kconfig_compiler.exe.

2) Adding "RENAME kconfig_compiler4" to the install(TARGETS ...) command 
does nothing.

3) Changing the install(TARGETS ...) command to

install(PROGRAMS 
${EXECUTABLE_OUTPUT_PATH}/kconfig_compiler${CMAKE_EXECUTABLE_SUFFIX}
EXPORT kdelibsToolsTargets RENAME 
kconfig_compiler4${CMAKE_EXECUTABLE_SUFFIX} DESTINATION ${BIN_INSTALL_DIR} )

causes a CMake Error: "install PROGRAMS given RENAME option with more 
than one file".  Placing RENAME before EXPORT causes a different error: 
"install PROGRAMS given unknown argument "EXPORT"".
(Continue reading)

ralf.habacker | 5 Jul 2009 21:41
Picon
Favicon
Gravatar

FW: KDE4 on cygwin, was: Re: KDE/kdelibs/cmake/modules

> Von: "Alexander Neundorf"

 
Hi Yaakov,

I got the email below from the kde build system
maintainer.

Regards

Ralf

> On Saturday 04 July 2009, Ralf Habacker wrote:
 > > Alexander Neundorf schrieb:
 > > > On Thursday 02 July 2009, Ralf Habacker wrote:
 > > >> SVN commit 990260 by habacker:
 > > >>
 > > >> Account for CMAKE_EXECUTABLE_SUFFIX in
KDE4_*_EXECUTABLE .shell
 > names,
 > > >> since that is how they are created. by cygwin ports
maintainer
 > > >
 > > > cygwin ?
 > >
 > > yes
 > >
 > > > Is anybody trying to build KDE4 for cygwin ?
 > >
 > > see
(Continue reading)

Dave Korn | 6 Jul 2009 03:28

Cygport big oopsie


  Ouch.  This one just bit me bigtime.  Spent the whole day wondering why
gcc-4.3.3-1 kept failing to build stage3 in libstdc++, which turned out to be
an autoconf test spuriously defining HAVE_FENV_H, which in turn turned out to
be because it wasn't passing the -nostdinc++ flag down on the command line
when preprocessing, which in turn turned out to be because ... well, after a
lot of debugging and wondering why it was configuring differently from a clean
build, it turns out to be because of a recent change in cygport.

> Revision 6684 - (view) (download) (annotate) - [select for diffs]
> Modified Fri Jun 5 01:28:24 2009 UTC (4 weeks, 2 days ago) by yselkowitz
> File length: 49059 byte(s)
> Diff to previous 6682
> 
> * bin/cygport.in: Export CC, CXX, and friends.
> * lib/cmake.cygclass: Use CC, CXX, and LDFLAGS.

 It was the new variable CPP in particular:

>  # Values which can be overridden either system-wide or per package
> +declare -x CC="gcc";
>  declare -x CFLAGS="-O2 -pipe ";
> +declare -x CPP="cpp";
>  declare -x CPPFLAGS="";
> +declare -x CXX="g++";
> +declare -x CXXCPP="cpp";
>  declare -x CXXFLAGS="${CFLAGS}";
> +declare -x F77="g77";
>  declare -x FFLAGS="${CFLAGS}";
> +declare -x GCJ="gcj";
(Continue reading)

Yaakov (Cygwin/X | 6 Jul 2009 03:57
Picon
Gravatar

Re: Cygport big oopsie

On 05/07/2009 20:28, Dave Korn wrote:
>> * bin/cygport.in: Export CC, CXX, and friends.
>> * lib/cmake.cygclass: Use CC, CXX, and LDFLAGS.

These were added, primarily, so that I could easily force gcc4 usage 
(with -shared-libgcc and -Wl,--enable-auto-import) for _all_ types of 
build systems.  I had hoped that we could make gcc4 the default earlier, 
but as that was rejected at the time, I maintain a local patch on top of 
this which does this for now.

>    If you declare CPP or any other such variable in the environment, it
> overrides any auto-detection that the package's configury might implement.
> That means that during a GCC bootstrap, it unconditionally uses "cpp" to
> invoke the preprocessor, when stages 2 and 3 very much want to use the
> previous stage's xgcc driver (along with assorted -nostdinc, -B, -isystem, and
> -E flags), and even stage 1 really needs to use "$(CC) -E".

Sounds like a bug in gcc machinery to me; if boostrapping, the supplied 
CC,CPP,... should only be accepted for stage1.  Does this affect CC as 
well during stages 2/3, or just CPP?

>    I don't think it's very safe to declare all this stuff in the environment
> when there are so many potential things that will behave one way or another
> based on a test for either existence or non-emptiness.  I think this is
> possibly a case of trying to be too helpful.  How about deleting the lot, or
> only declaring them for cmake builds, or ... any other ideas?

The best solution would be to have a stable, preferred-by-alternatives 
gcc4 with -shared-libgcc on by default, in which case I can stop 
overriding these.  (--enable-auto-import is on by default now with the 
(Continue reading)

Dave Korn | 6 Jul 2009 06:29

Re: Cygport big oopsie

Yaakov (Cygwin/X) wrote:
> On 05/07/2009 20:28, Dave Korn wrote:

>>    If you declare CPP or any other such variable in the environment, it
>> overrides any auto-detection that the package's configury might implement.
>> That means that during a GCC bootstrap, it unconditionally uses "cpp" to
>> invoke the preprocessor, when stages 2 and 3 very much want to use the
>> previous stage's xgcc driver (along with assorted -nostdinc, -B, -isystem, and
>> -E flags), and even stage 1 really needs to use "$(CC) -E".
> 
> Sounds like a bug in gcc machinery to me; if boostrapping, the supplied 
> CC,CPP,... should only be accepted for stage1.  Does this affect CC as 
> well during stages 2/3, or just CPP?

  Just CPP, as far as I noticed.  It could well be a bug.

> (--enable-auto-import is on by default now with the 
> new binutils-2.19.51 release, right?)

  Yep :)

>> I can unset all these vars in my project's .cygport,
> 
> Probably not a bad idea in any case, given the nature of the gcc build; 
> you never know what someone might have set in their environment.
> 
>> but isn't doing this liable to have unexpected effects all over the place?
> 
> Actually it's been quite helpful, but then again, I'm not building gcc.
> 
(Continue reading)


Gmane