Bug#972912: [Pkg-samba-maint] Bug#972912: Incorrect fix to bug

2021-01-08 Thread Sven Mueller
Ack, merge request created.
I had already built ldb with the patch for both unstable and the
environment I ran into issues due to the symbols check and also rebuilt
samba in both afterwards. Both built fine and at least basic functionality
tests also worked fine.

Am Fr., 8. Jan. 2021 um 13:33 Uhr schrieb Mathieu Parent <
math.par...@gmail.com>:

> Le lun. 21 déc. 2020 à 16:06, Sven Mueller  a
> écrit :
> >
> > Well, turns out that apparently nobody else bothers with a .symbols file
> for Python extensions. I looked at the packages for samba, numpy, mypy and
> python-stdlib-extensions. And if the Python maintainers themselves don't do
> it, you probably shouldn't.
> >
> > I attached a diff to remove the relevant file and the setup for it. I
> also added some verbosity to the stuff debhelper does (I find it harder to
> debug build issues without it).
> > I verified that except for this symbols file going away, nothing else
> changed. (Most notable, the main libldb-dev package still looks the same.)
> - Verified via diffoscope which only showed expected changes (timestamps,
> mostly)
> >
> > I'll see if I can turn it into a pull request on Salsa, but my git-foo
> is weaker than it probably should be, so feel free to just apply my patch
> yourself.
> > If I create a pull request, should that include an update to
> debian/changelog?
>
> Hello, just a quick note that I won't work on this for bullseye. If
> you can propose a MR in salsa and test that samba builds and works,
> I'll happily review it.
>
> Regards
> --
> Mathieu Parent
>


Bug#972912: [Pkg-samba-maint] Bug#972912: Incorrect fix to bug

2021-01-08 Thread Mathieu Parent
Le lun. 21 déc. 2020 à 16:06, Sven Mueller  a écrit :
>
> Well, turns out that apparently nobody else bothers with a .symbols file for 
> Python extensions. I looked at the packages for samba, numpy, mypy and 
> python-stdlib-extensions. And if the Python maintainers themselves don't do 
> it, you probably shouldn't.
>
> I attached a diff to remove the relevant file and the setup for it. I also 
> added some verbosity to the stuff debhelper does (I find it harder to debug 
> build issues without it).
> I verified that except for this symbols file going away, nothing else 
> changed. (Most notable, the main libldb-dev package still looks the same.) - 
> Verified via diffoscope which only showed expected changes (timestamps, 
> mostly)
>
> I'll see if I can turn it into a pull request on Salsa, but my git-foo is 
> weaker than it probably should be, so feel free to just apply my patch 
> yourself.
> If I create a pull request, should that include an update to debian/changelog?

Hello, just a quick note that I won't work on this for bullseye. If
you can propose a MR in salsa and test that samba builds and works,
I'll happily review it.

Regards
-- 
Mathieu Parent



Bug#972912: [Pkg-samba-maint] Bug#972912: Incorrect fix to bug

2020-12-21 Thread Sven Mueller
Well, turns out that apparently nobody else bothers with a .symbols file
for Python extensions. I looked at the packages for samba, numpy, mypy and
python-stdlib-extensions. And if the Python maintainers themselves don't do
it, you probably shouldn't.

I attached a diff to remove the relevant file and the setup for it. I also
added some verbosity to the stuff debhelper does (I find it harder to debug
build issues without it).
I verified that except for this symbols file going away, nothing else
changed. (Most notable, the main libldb-dev package still looks the same.)
- Verified via diffoscope which only showed expected changes (timestamps,
mostly)

I'll see if I can turn it into a pull request on Salsa, but my git-foo
is weaker than it probably should be, so feel free to just apply my patch
yourself.
If I create a pull request, should that include an update to
debian/changelog?

Cheers,
Sven


Am Do., 17. Dez. 2020 um 16:55 Uhr schrieb Mathieu Parent <
math.par...@gmail.com>:

> Le jeu. 17 déc. 2020 à 16:48, Sven Mueller  a
> écrit :
> >
> > Hi Mathieu.
>
> Hi,
>
> > Just wanted to say that your fix here seems wrong:
> > The symbols file says when a specific symbol for a specific lib was
> added.
> > If I rebuild ldb against Python 3.9, it will suddenly claim that - for
> example -  symbol PYLDB_UTIL_2.1.0@PYLDB_UTIL_2.1.0 was added to the
> package - for the Python 3.9 specific lib - in package version 2:2.1.0 -
> Even though that package version was not built against Python 3.9 at all.
> >
> > The better fix would be to explicitly build against specific Python
> versions (python3.8-dev, python3.9-dev build dependencies) and have
> appropriate symbols listed for both of them.
> >
> > Currently, if a package builds against the ldb python bindings for
> Python 3.9, it will generate versioned dependencies that are incorrect (if
> all it uses would be the above symbol, it would depend on python3-ldb >=
> 2:2.1.0 - which didn't have any Python 3.9 bindings) - and fail after
> installation.
> >
> > To be fair though, I'm not even sure having the symbols file for the
> python bindings .so files makes much sense.
>
> OK. Could you submit a merge request fixing this? SInce the migration
> to python3, the bindings are getting complicated. Any help here is
> apprecciated.
>
> Regards
>
> Mathieu Parent
>
diff -urN ldb-2.2.0.orig/debian/python3-ldb.symbols.in ldb-2.2.0/debian/python3-ldb.symbols.in
--- ldb-2.2.0.orig/debian/python3-ldb.symbols.in	2020-12-21 13:23:55.062566080 +0100
+++ ldb-2.2.0/debian/python3-ldb.symbols.in	1970-01-01 01:00:00.0 +0100
@@ -1,63 +0,0 @@
-#!/usr/bin/dh-exec
-libpyldb-util${DEB_PY3_EXTENSION_SUFFIX}.2 python3-ldb #MINVER#
- PYLDB_UTIL${DEB_PY3_EXTENSION_UPCASE}_2.2.0@PYLDB_UTIL${DEB_PY3_EXTENSION_UPCASE}_2.2.0 2:2.2.0
-#include "python3-ldb.symbols.common" PYLDB_UTIL_1.1.2@PYLDB_UTIL_1.1.2 2:2.0.7
- PYLDB_UTIL_1.1.2@PYLDB_UTIL_1.1.2 2:2.2.0
- PYLDB_UTIL_1.1.3@PYLDB_UTIL_1.1.3 2:2.0.7
- PYLDB_UTIL_1.1.4@PYLDB_UTIL_1.1.4 2:2.0.7
- PYLDB_UTIL_1.1.5@PYLDB_UTIL_1.1.5 2:2.0.7
- PYLDB_UTIL_1.1.6@PYLDB_UTIL_1.1.6 2:2.0.7
- PYLDB_UTIL_1.1.7@PYLDB_UTIL_1.1.7 2:2.0.7
- PYLDB_UTIL_1.1.8@PYLDB_UTIL_1.1.8 2:2.0.7
- PYLDB_UTIL_1.1.9@PYLDB_UTIL_1.1.9 2:2.0.7
- PYLDB_UTIL_1.1.10@PYLDB_UTIL_1.1.10 2:2.0.7
- PYLDB_UTIL_1.1.11@PYLDB_UTIL_1.1.11 2:2.0.7
- PYLDB_UTIL_1.1.12@PYLDB_UTIL_1.1.12 2:2.0.7
- PYLDB_UTIL_1.1.13@PYLDB_UTIL_1.1.13 2:2.0.7
- PYLDB_UTIL_1.1.14@PYLDB_UTIL_1.1.14 2:2.0.7
- PYLDB_UTIL_1.1.15@PYLDB_UTIL_1.1.15 2:2.0.7
- PYLDB_UTIL_1.1.16@PYLDB_UTIL_1.1.16 2:2.0.7
- PYLDB_UTIL_1.1.17@PYLDB_UTIL_1.1.17 2:2.0.7
- PYLDB_UTIL_1.1.18@PYLDB_UTIL_1.1.18 2:2.0.7
- PYLDB_UTIL_1.1.19@PYLDB_UTIL_1.1.19 2:2.0.7
- PYLDB_UTIL_1.1.20@PYLDB_UTIL_1.1.20 2:2.0.7
- PYLDB_UTIL_1.1.21@PYLDB_UTIL_1.1.21 2:2.0.7
- PYLDB_UTIL_1.1.22@PYLDB_UTIL_1.1.22 2:2.0.7
- PYLDB_UTIL_1.1.23@PYLDB_UTIL_1.1.23 1.5.4
- PYLDB_UTIL_1.1.24@PYLDB_UTIL_1.1.24 1.5.4
- PYLDB_UTIL_1.1.25@PYLDB_UTIL_1.1.25 1.5.4
- PYLDB_UTIL_1.1.26@PYLDB_UTIL_1.1.26 1.5.4
- PYLDB_UTIL_1.1.27@PYLDB_UTIL_1.1.27 1.5.4
- PYLDB_UTIL_1.1.28@PYLDB_UTIL_1.1.28 1.5.4
- PYLDB_UTIL_1.1.29@PYLDB_UTIL_1.1.29 1.5.4
- PYLDB_UTIL_1.1.30@PYLDB_UTIL_1.1.30 1.5.4
- PYLDB_UTIL_1.1.31@PYLDB_UTIL_1.1.31 1.5.4
- PYLDB_UTIL_1.2.0@PYLDB_UTIL_1.2.0 1.5.4
- PYLDB_UTIL_1.2.1@PYLDB_UTIL_1.2.1 1.5.4
- PYLDB_UTIL_1.2.2@PYLDB_UTIL_1.2.2 1.5.4
- PYLDB_UTIL_1.2.3@PYLDB_UTIL_1.2.3 1.5.4
- PYLDB_UTIL_1.3.0@PYLDB_UTIL_1.3.0 1.5.4
- PYLDB_UTIL_1.3.1@PYLDB_UTIL_1.3.1 1.5.4
- PYLDB_UTIL_1.3.2@PYLDB_UTIL_1.3.2 1.5.4
- PYLDB_UTIL_1.4.0@PYLDB_UTIL_1.4.0 1.5.4
- PYLDB_UTIL_1.4.1@PYLDB_UTIL_1.4.1 1.5.4
- PYLDB_UTIL_1.5.0@PYLDB_UTIL_1.5.0 1.5.4
- PYLDB_UTIL_1.5.1@PYLDB_UTIL_1.5.1 1.5.4
- PYLDB_UTIL_1.5.2@PYLDB_UTIL_1.5.2 1.5.4
- PYLDB_UTIL_1.5.3@PYLDB_UTIL_1.5.3 1.5.4
- PYLDB_UTIL_1.6.0@PYLDB_UTIL_1.6.0 2:2.0.7
- PYLDB_UTIL_1.6.1@PYLDB_UTIL_1.6.1 2:2.0.7
- PYLDB_UTIL_1.6.2@PYLDB_UTIL_1.6.2 2:2.0.7
- PYLDB_UTIL_1.6.3@PYLDB_UTIL_1.6.3 2:2.0.7
- PYLDB_UTIL_2.0.0@PYLDB_UTIL_2.0.0 2:2.0.7
- 

Bug#972912: [Pkg-samba-maint] Bug#972912: Incorrect fix to bug

2020-12-17 Thread Mathieu Parent
Le jeu. 17 déc. 2020 à 16:48, Sven Mueller  a écrit :
>
> Hi Mathieu.

Hi,

> Just wanted to say that your fix here seems wrong:
> The symbols file says when a specific symbol for a specific lib was added.
> If I rebuild ldb against Python 3.9, it will suddenly claim that - for 
> example -  symbol PYLDB_UTIL_2.1.0@PYLDB_UTIL_2.1.0 was added to the package 
> - for the Python 3.9 specific lib - in package version 2:2.1.0 - Even though 
> that package version was not built against Python 3.9 at all.
>
> The better fix would be to explicitly build against specific Python versions 
> (python3.8-dev, python3.9-dev build dependencies) and have appropriate 
> symbols listed for both of them.
>
> Currently, if a package builds against the ldb python bindings for Python 
> 3.9, it will generate versioned dependencies that are incorrect (if all it 
> uses would be the above symbol, it would depend on python3-ldb >= 2:2.1.0 - 
> which didn't have any Python 3.9 bindings) - and fail after installation.
>
> To be fair though, I'm not even sure having the symbols file for the python 
> bindings .so files makes much sense.

OK. Could you submit a merge request fixing this? SInce the migration
to python3, the bindings are getting complicated. Any help here is
apprecciated.

Regards

Mathieu Parent