Bugzilla Bug 12081

Hi !

This new library provides Ada95 XML support.

Package requires either compiled Ada language support in the gcc-3.2 or
separate gnat compiler.

For more information:
http://libre.act-europe.fr/xmlada

Created an attachment (id=6466) [edit]
xmlada-0.7.1.ebuild

Created an attachment (id=6467) [edit]
Patch to increase the stack size

Created an attachment (id=6468) [edit]
the first ChangeLog

Can't get around to this anytime soon, I'm afraid.

Hi Alexis.

Thanks a lot for the submission!
I have cleaned up the ebuild and committed it. Please test.
Few notes:
1. I added all the necessary formatting requisites. Issue of note: please make sure your ebuilds have a proper header (other staff wasn't defined at the time of submission, but now ebuilds also need to have KEYWORDS, SLOT and bunch of other stuff as in skel.ebuild..)

2. I used gnat.eclass in order to do some common stuff and strip inappropriate flags.

3. I removed the patch and replaced it with a simple sed instruction

4. !!The docs could not be built - tex was reporting some problems and aborting. I have removed a target for building dvi. All other forms are getting installed, although you might look into the issue if you would like to get dvi built.

5. doc installation has been redone, please take a look at the ebuild.

George

src_install() has to be changed to look like this in order for the package to
follow the GNAE standard:

src_install ()
{
    dodoc docs/*.{html,ps,txt}
    doinfo docs/*.info

    make prefix=${D}/usr \
        libdir=${D}/usr/lib/ada/adalib \
        incdir=${D}/usr/lib/ada/adainclude/xmlada \
        alidir=${D}/usr/lib/ada/adainclude/xmlada install \
        || die "install failed"
}

*** Bug 25065 has been marked as a duplicate of this bug. ***

Hi David.

Thanks for taking a look at this one, I suspected something like that.
Unfortunately the fix you propose results i sandbox violations:

cd obj; cp -p -f *.ali /usr/include/xmlada
ACCESS DENIED  open_wr:   /usr/include/xmlada/unicode.ali
ACCESS DENIED  unlink:    /usr/include/xmlada/unicode.ali
...

and the whole bunch like that..
Instead I just moved the files around manually in the -r1 version of ebuild.
Please check (the ebuild is committed).

George

Hello,

The src_unpack() has to be changed by adding these lines :

sed -i -e "s#${prefix}/lib#${prefix}/lib/ada/adalib/xmlada#" xmlada-config.in
sed -i -e "s#${prefix}/include/xmlada#${prefix}/lib/ada/adainclude/xmlada#" xmlada-config.in

To adapt xmlada-config script with the GNAE installation.

Alexis

Hi guys.

Thanks for the fix and suggestions!
Alexis: I added these lines to src_unpack
David:  I fixed where ali's are out, sorry for improper positioning of them originally - I was trying to model it after other stuff which you already GNAE-ceaned, but apparently missed this one.

The ebuild is updated, please test.

George

It's probably my fault, some of my initial ebuilds installed the alis in
adainclude. Luckily none of those were commited. Also in the src_install()
sample I posted here the alis are installed in adainclude which is probably why
I hadn't posted it yet.
Your new xmlada seems perfect now though.
We really need to add some policy stuff regarding Ada ebuilds if we want users
to create GNAE-compatible ebuilds from the start, the question is where?

-m3dnow should probably be added to the stripped flags as well. What about if a
user uses -mfpmath=sse,387 or -mfpmath=387? Is it possible to add something
like -mfpmath=* to filter-flags or would you have to add one entry for every
case?

Please direct any testreports to the -r2 bugreport. 

Tested -r2, seems fine.
Closing the bug (once again :)).

George