GtkAda on Windows: Does anyone have a gtkada.gpr file?

GtkAda on Windows: Does anyone have a gtkada.gpr file?

[Michael R]

Hi,

I've just starting to look at GtkAda on Windows.  My Ubuntu system
makes it relatively
easy by supplying a gtkada2.gpr project file.  The standard GtkAda
distribution does
not supply a project file.  The documentation suggest simply add

   -IC:\GtkAda\include\gtkada

to a "gnatmake" command line is all that is needed.  This is failing
for the simple.adb
example:

   > gnatmake simple -IC:\GtkAda\include\gtkada
   gcc -c -IC:\GtkAda\include\gtkada simple.adb
   gnatbind -IC:\GtkAda\include\gtkada -x simple.ali
   gnatlink simple.ali
   b~simple.o:b~simple.adb:(.text+0x225): undefined reference to
`glib___elabs'
   b~simple.o:b~simple.adb:(.text+0x22b): undefined reference to
`glib_E'
   b~simple.o:b~simple.adb:(.text+0x231): undefined reference to
`gtkada__types___elabs'
   b~simple.o:b~simple.adb:(.text+0x237): undefined reference to
`gtkada__types_E'
   b~simple.o:b~simple.adb:(.text+0x23e): undefined reference to
`gdk__rectangle_E'
   b~simple.o:b~simple.adb:(.text+0x245): undefined reference to
`glib__glist_E'
   ...

I guess I'm obviously missing something here.  Anyone have any
suggestions?

Take care,
Michael.

Re: GtkAda on Windows: Does anyone have a gtkada.gpr file?

[Rick]

Check out
http://adasafehouse.webs.com/GtkAda.html#Getting Going

I think all you will need is there.

Re: GtkAda on Windows: Does anyone have a gtkada.gpr file?

[Dmitry A. Kazakov]

On Sun, 17 Oct 2010 14:32:25 -0700 (PDT), Michael R wrote:

> I've just starting to look at GtkAda on Windows.  My Ubuntu system
> makes it relatively
> easy by supplying a gtkada2.gpr project file.

Yep, instead of gtkada.gpr they invented gtkada2.gpr. When I use Fedora's
GtkAda I add my own gtkada.gpr and some other gpr files to keep it
consistent.

But since GNAT FSF for Fedora/Ubuntu is quite broken anyway, I would not
worry about that in your place. You better use GNAT GPL 2010 for Linux and
its GtkAda instead. BTW, the Fedora GtkAda is also outdated, when I checked
it last time. In particular pixbuf was not an gobject, if I correctly
remember. That is very nasty incompatibility. In short, see above, do not
use it.

> The standard GtkAda distribution does
> not supply a project file.

No it does. That is C:\GtkAda\lib\gnat

> The documentation suggest simply add
> 
>    -IC:\GtkAda\include\gtkada
> 
> to a "gnatmake" command line is all that is needed.  This is failing
> for the simple.adb example:

[..link errors...]

> I guess I'm obviously missing something here.  Anyone have any
> suggestions?

1. If you are going to link using naked gnatmake, you have to add options
to the linker, like -lgtk-win32-2.0 (and whatever, the list is long).

2. If you want to use gtkada.gpr, which is preferable of course, you need
to make it available. There are many ways. One is to integrate it into the
GNAT, like win32ada is. That is a bit complicated. A simpler one is to just
add C:\GtkAda\lib\gnat to the ADA_PROJECT_PATH environment variable.

-- 
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de

Re: GtkAda on Windows: Does anyone have a gtkada.gpr file?

[Ludovic Brenta]

Dmitry A. Kazakov wrote on comp.lang.ada:
> On Sun, 17 Oct 2010 14:32:25 -0700 (PDT), Michael R wrote:
>> I've just starting to look at GtkAda on Windows.  My Ubuntu system
>> makes it relatively easy by supplying a gtkada2.gpr project file.
>
> Yep, instead of gtkada.gpr they invented gtkada2.gpr.

"They" were me and the reason was that I wanted to support both GtkAda
1.x and 2.x on the same machine; that never happened for lack of
manpower.

> When I use Fedora's GtkAda I add my own gtkada.gpr and some other gpr
> files to keep it consistent.

The OP never mentioned Fedora. In case you really meant Debian, I
remind your that per your request, gtkada.gpr was reinstated in Debian
starting with libgtkada2.14.2-dev in June this year. This is present
in Debian testing and probably in Ubuntu 10.10.

> But since GNAT FSF for Fedora/Ubuntu is quite broken anyway,

The GNAT in Debian (and hence Ubuntu) is far from broken and it works
quite well for me. I think your remark is a bit offensive given all
the hard work I put in the Debian compiler. In particular I backported
a lot of bug fixes into gnat-4.4, some even at your request (e.g.
http://bugs.debian.org/589164). Unless Fedora has the same level of
support and the same backported fixes, you should not imply that the
Debian and Fedora compilers are identical.

> I would not worry about that in your place.

The point of a binary distribution like Debian and Ubuntu is to take
the installation, compilation and configuration worries away. The OP
does not worry about his Ubuntu, he worries about GNAT GPL on Windows.
So your advice is misdirected.

> You better use GNAT GPL 2010 for
> Linux and its GtkAda instead. BTW, the Fedora GtkAda is also outdated, when
> I checked it last time. In particular pixbuf was not an gobject, if I correctly
> remember. That is very nasty incompatibility. In short, see above, do not
> use it.

The OP never mentioned Fedora and he is not using it, so you should be
happy :)

--
Ludovic Brenta.

Re: GtkAda on Windows: Does anyone have a gtkada.gpr file?

[Georg Bauhaus]

On 18.10.10 14:43, Ludovic Brenta wrote:
> Dmitry A. Kazakov wrote on comp.lang.ada:

>> But since GNAT FSF for Fedora/Ubuntu is quite broken anyway,
> 
> The GNAT in Debian (and hence Ubuntu) is far from broken and it works
> quite well for me. I think your remark is a bit offensive given all
> the hard work I put in the Debian compiler. In particular I backported
> a lot of bug fixes into gnat-4.4, some even at your request (e.g.
> http://bugs.debian.org/589164).

I'm not sure, looking at The Computer Language Benchmark Game.

Do you, by any chance, know the state of gnat-4.4 in Debian vs Ubuntu?
The Ada ranking in the Shootout benchmark appears to be suffering from
gcc-4.4 in Ubuntu. The Mandelbrot program, for example, once ran fine
and fast on all kinds of GNAT supported platforms.  It has now been replaced
with its precursor (which runs more slowly), since the executable
produced by gcc-4.4 gives bad output.

This isn't the first time that GCC's GNAT part appears to be behind the
C and C++ frontends, in terms of quality.  This spoils some efforts.
What needs to be done?


Georg

Re: GtkAda on Windows: Does anyone have a gtkada.gpr file?

[Dmitry A. Kazakov]

On Mon, 18 Oct 2010 05:43:39 -0700 (PDT), Ludovic Brenta wrote:

> Dmitry A. Kazakov wrote on comp.lang.ada:
>> On Sun, 17 Oct 2010 14:32:25 -0700 (PDT), Michael R wrote:
>>> I've just starting to look at GtkAda on Windows. �My Ubuntu system
>>> makes it relatively easy by supplying a gtkada2.gpr project file.
>>
>> Yep, instead of gtkada.gpr they invented gtkada2.gpr.
> 
> "They" were me and the reason was that I wanted to support both GtkAda
> 1.x and 2.x on the same machine; that never happened for lack of
> manpower.

So it was you! (:-))

There was no need to introduce two gpr files as you could switch between
versions within one gtkada.gpr using a scenario variable. This is how I
deal with incompatible GtkAda changes made between 2.14.0 and 2.14.2.

> I think your remark is a bit offensive given all
> the hard work I put in the Debian compiler.

I didn't mean to offend anybody, just pointed out a potential problem.

Of course I appreciate you and all others who are fixing the compiler. That
is without a question.

> In particular I backported
> a lot of bug fixes into gnat-4.4, some even at your request (e.g.
> http://bugs.debian.org/589164). Unless Fedora has the same level of
> support and the same backported fixes, you should not imply that the
> Debian and Fedora compilers are identical.

I don't. But Debian GNAT is still broken. I checked it two days ago under
fully updated squeeze. Controlled types are broken.

Sorry for misunderstanding. I had an impression that the OP wanted to
design a cross-platform GtkAda-based application. Since I had spent
(wasted) a lot of time on this stuff I wished to warn him. This won't work
with the FSF GNAT, because of compiler bugs, because of incompatible GtkAda
distributions. Use GNAT GPL 2010 and patiently wait until FSF catches up.

-- 
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de

Re: GtkAda on Windows: Does anyone have a gtkada.gpr file?

[Ludovic Brenta]

Georg Bauhaus wrote on comp.lang.ada:
> On 18.10.10 14:43, Ludovic Brenta wrote:
>> Dmitry A. Kazakov wrote on comp.lang.ada:
>>> But since GNAT FSF for Fedora/Ubuntu is quite broken anyway,
>
>> The GNAT in Debian (and hence Ubuntu) is far from broken and it works
>> quite well for me. I think your remark is a bit offensive given all
>> the hard work I put in the Debian compiler. In particular I backported
>> a lot of bug fixes into gnat-4.4, some even at your request (e.g.
>> http://bugs.debian.org/589164).
>
> I'm not sure, looking at The Computer Language Benchmark Game.
>
> Do you, by any chance, know the state of gnat-4.4 in Debian vs Ubuntu?

I don't follow Ubuntu; just keeping up with Debian is enough work :)
But in theory, they should be quite close as the same person is the
maintainer of GCC in both Debian and Ubuntu.  He also maintains GNAT
in Ubuntu.  Now it can be that other packages like the kernel and
glibc, or their configuration, introduce subtle differences between
Debian and Ubuntu.

> The Ada ranking in the Shootout benchmark appears to be suffering from
> gcc-4.4 in Ubuntu. The Mandelbrot program, for example, once ran fine
> and fast on all kinds of GNAT supported platforms.  It has now been replaced
> with its precursor (which runs more slowly), since the executable
> produced by gcc-4.4 gives bad output.
>
> This isn't the first time that GCC's GNAT part appears to be behind the
> C and C++ frontends, in terms of quality.  This spoils some efforts.
> What needs to be done?

Report problems through the Debian bug tracking system as precisely as
possible.
If you discover a bug in gnat-4.4 which happens not to be present in
GCC 4.5 or 4.6, I'll try hard to backport the fix.

As I've said many times before, not reporting a bug is the surest way
to never see it fixed.

--
Ludovic Brenta.

Re: GtkAda on Windows: Does anyone have a gtkada.gpr file?

[Ludovic Brenta]

Dmitry A. Kazakov wrote on comp.lang.ada:
> [...] Debian GNAT is still broken. I checked it two days ago under
> fully updated squeeze. Controlled types are broken.

I am interested in precise bug reports; vague claims do not help
anyone even if they are true.

You seem to assume that AdaCore will learn of your bugs by chance or
magic and fix them in the next GNAT GPL Edition.  Even if that were
true, allowing this to happen makes you more and more dependent on
AdaCore.  In contrast, reporting bugs in the Debian bug tracking
system or the GCC Bugzilla raises awareness of the Ada part of GCC,
motivates more people to contribute to it and initiates a positive
feedback loop:

More users -> more bug reports -> more contributors -> more fixes ->
more users.

Waiting patiently for the next GNAT GPL Edition has its virtues but
helping the advancement of the compiler is not one of them.

--
Ludovic Brenta.

Re: GtkAda on Windows: Does anyone have a gtkada.gpr file?

[Dmitry A. Kazakov]

On Mon, 18 Oct 2010 08:40:11 -0700 (PDT), Ludovic Brenta wrote:

> Dmitry A. Kazakov wrote on comp.lang.ada:
>> [...] Debian GNAT is still broken. I checked it two days ago under
>> fully updated squeeze. Controlled types are broken.
> 
> I am interested in precise bug reports; vague claims do not help
> anyone even if they are true.

There is a test for this bug in:

   http://www.dmitry-kazakov.de/ada/components.htm

   .../test_components/test_handles.adb

the project file is

   .../test_components/components-tests.gpr

it fails with Program_Error when something goes wrong with controlled
type's Initialize/Finalize/Adjust calls.

> You seem to assume that AdaCore will learn of your bugs by chance or
> magic and fix them in the next GNAT GPL Edition.

There is no magic. I report bugs to GNAT Pro and it gets fixed in the GPL
automatically as AdaCore updates the GPL. AFAIK GPL is a mere snapshot of
Pro.

This particular bug was two or so years ago fixed in GNAT Pro. Later it was
in GNAT GPL 2009, or maybe 2010 guess, it is certainly absent in GNAT GPL
2010. Would FSF be synched to GPL 2010, the bug should disappear.

-- 
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de

Re: GtkAda on Windows: Does anyone have a gtkada.gpr file?

[Ludovic Brenta]

Dmitry A. Kazakov wrote on comp.lang.ada:
> On Mon, 18 Oct 2010 08:40:11 -0700 (PDT), Ludovic Brenta wrote:
> > Dmitry A. Kazakov wrote on comp.lang.ada:
>>> [...] Debian GNAT is still broken. I checked it two days ago under
>>> fully updated squeeze. Controlled types are broken.
>
>> I am interested in precise bug reports; vague claims do not help
>> anyone even if they are true.
>
> There is a test for this bug in:
>
>    http://www.dmitry-kazakov.de/ada/components.htm
>
>    .../test_components/test_handles.adb
>
> the project file is
>
>    .../test_components/components-tests.gpr
>
> it fails with Program_Error when something goes wrong with controlled
> type's Initialize/Finalize/Adjust calls.
>
>> You seem to assume that AdaCore will learn of your bugs by chance or
>> magic and fix them in the next GNAT GPL Edition.
>
> There is no magic. I report bugs to GNAT Pro and it gets fixed in the GPL
> automatically as AdaCore updates the GPL. AFAIK GPL is a mere snapshot of
> Pro.
>
> This particular bug was two or so years ago fixed in GNAT Pro. Later it was
> in GNAT GPL 2009, or maybe 2010 guess, it is certainly absent in GNAT GPL
> 2010. Would FSF be synched to GPL 2010, the bug should disappear.

FSF GCC does receive all fixes from GNAT Pro/GPL (in the trunk) but
the stable releases can be months behind the GPL release. Chances are
high that this particular bug is already fixed in GCC  4.5 or the
future 4.6 (i.e. trunk). If so, bisecting and backporting to GCC 4.4
should be straightforward.

--
Ludovic Brenta.

Re: GtkAda on Windows: Does anyone have a gtkada.gpr file?

[Simon Wright]

"Dmitry A. Kazakov" <mailbox@dmitry-kazakov.de> writes:

> There is a test for this bug in:
>
>    http://www.dmitry-kazakov.de/ada/components.htm
>
>    .../test_components/test_handles.adb
>
> the project file is
>
>    .../test_components/components-tests.gpr
>
> it fails with Program_Error when something goes wrong with controlled
> type's Initialize/Finalize/Adjust calls.

Fixed in GCC 4.5.0.

Bug in gnat-4.4 (was: GtkAda on Windows: Does anyone have a gtkada.gpr file?)

[Ludovic Brenta]

Dmitry A. Kazakov writes on comp.lang.ada:
> On Mon, 18 Oct 2010 08:40:11 -0700 (PDT), Ludovic Brenta wrote:
>
>> Dmitry A. Kazakov wrote on comp.lang.ada:
>>> [...] Debian GNAT is still broken. I checked it two days ago under
>>> fully updated squeeze. Controlled types are broken.
>> 
>> I am interested in precise bug reports; vague claims do not help
>> anyone even if they are true.
>
> There is a test for this bug in:
>
>    http://www.dmitry-kazakov.de/ada/components.htm
>
>    .../test_components/test_handles.adb
>
> the project file is
>
>    .../test_components/components-tests.gpr
>
> it fails with Program_Error when something goes wrong with controlled
> type's Initialize/Finalize/Adjust calls.

$ dpkg -l gnat-4.4 |grep ^i
ii  gnat-4.4                4.4.4-5          The GNU Ada compiler
$ mkdir components-3.9; cd components-3.9
$ tar xzf ~/incoming/components_3_9.tgz
$ gnatmake -Ptest_components/components-tests.gpr test_handles.adb
$ test_components/test_handles
Testing handles and sets ...
... Done
$

No Program_Error. Are you sure this is the correct way to reproduce your
bug?

-- 
Ludovic Brenta.

Re: Bug in gnat-4.4

[Dmitry A. Kazakov]

On Mon, 18 Oct 2010 21:41:56 +0200, Ludovic Brenta wrote:

> No Program_Error.

Good news!

> Are you sure this is the correct way to reproduce your
> bug?

Yes, it is. But I have gnat-4.4.4-4. apt-get update/upgrade keeps it old.
I'll try aptitude dist-upgrade. How wonderful it is to have a dozen
utilities for doing the same thing...

-- 
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de

Re: Bug in gnat-4.4

[Ludovic Brenta]

"Dmitry A. Kazakov" <mailbox@dmitry-kazakov.de> writes:
> On Mon, 18 Oct 2010 21:41:56 +0200, Ludovic Brenta wrote:
>> No Program_Error.
>
> Good news!
>
>> Are you sure this is the correct way to reproduce your bug?
>
> Yes, it is. But I have gnat-4.4.4-4.

That means your bug was really #589164, as I suspected.  The backport
for the fix is in -5.

> apt-get update/upgrade keeps it old.  I'll try aptitude
> dist-upgrade. How wonderful it is to have a dozen utilities for doing
> the same thing...

Since all these utilities use the same apt back-end, they all give the
same results.  What do you have in /etc/apt/sources.list?

-- 
Ludovic Brenta.

Re: Bug in gnat-4.4

[Michael R]

On Oct 18, 1:38 pm, Ludovic Brenta <ludo...@ludovic-brenta.org> wrote:
> "Dmitry A. Kazakov" <mail...@dmitry-kazakov.de> writes:
>
> > On Mon, 18 Oct 2010 21:41:56 +0200, Ludovic Brenta wrote:
> >> No Program_Error.
>
> > Good news!
>
> >> Are you sure this is the correct way to reproduce your bug?
>
> > Yes, it is. But I have gnat-4.4.4-4.
>
> That means your bug was really #589164, as I suspected.  The backport
> for the fix is in -5.
>
> > apt-get update/upgrade keeps it old.  I'll try aptitude
> > dist-upgrade. How wonderful it is to have a dozen utilities for doing
> > the same thing...
>
> Since all these utilities use the same apt back-end, they all give the
> same results.  What do you have in /etc/apt/sources.list?
>
> --
> Ludovic Brenta.

Hi Folks,

Thank you for the feedback.  I had originally dismissed the gtkada.gpr
on Windows
as it seemed, on first glance, to be a Unix file.  Using the as
distributed file
works OK.

I've switched to simply using "gtkada.gpr" rather than "gtkada2.gpr"
and am manually
renaming on Ubuntu just to have a consistent view of the world.

My Gtk app was just a test app but I wanted to verify it on Windows.

Take care,
Michael

Re: Bug in gnat-4.4

[Dmitry A. Kazakov]

On Mon, 18 Oct 2010 22:38:47 +0200, Ludovic Brenta wrote:

> Since all these utilities use the same apt back-end, they all give the
> same results.  What do you have in /etc/apt/sources.list?

Never looked in, but aptitude dist-upgrade did the trick. I think it was
some dependencies preventing removal of old packages. Anyway, the bug is
finally out in 4.4.4-5. Congratulations!

I will test 4.4.4-5 for other problems (there were compiler crashes upon
instantiation of generics etc).

-- 
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de

Bug in gnat-4.4, revisited

[Dmitry A. Kazakov]

On Mon, 18 Oct 2010 22:38:47 +0200, Ludovic Brenta wrote:

> That means your bug was really #589164, as I suspected.  The backport
> for the fix is in -5.

OK, I have performed some further checks.

It seems that some traces of the bug are still present in 4.4.4-5. It is
more complicated now. test_handles works, but another test fails. When it
is compiled this way:

$ gcc -c sqlite-sources/sqlite3.c
$ gnatmake -O2 test_sqlite_persistence.adb -largs sqlite3.o -ldl

then

$ ./test_sqlite_persistence
Session 1
\_1
  |\_2
  |  |\_*
  |   \_*
   \_3
     |\_4
     |  |\_5
     |  |  |\_*
     |  |   \_*
     |   \_*
      \_6
        |\_*
         \_*

raised PROGRAM_ERROR : adjust/finalize raised PROGRAM_ERROR:
PERSISTENT.SQLITE.DATA_BASE_OBJECT is still in use

Interestingly, when compiled without -O2:

$ gnatmake test_sqlite_persistence.adb -largs sqlite3.o -ldl

the test does not fail.

It seems that the fix does not completely eliminate the bug when
optimization is on.

-- 
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de