rpath and /usr/lib/package directories

2006-07-15 Thread Charles Fry
Hi,

I recently uploaded courierpassd, which wraps some functionality of
courier-authlib in the popassd protocol. lintian has been warning me
about the use of rpath, about which I posted on debian-mentors. This
culminated in the following thread:

http://lists.debian.org/debian-mentors/2006/07/msg00221.html

Which along with several other related comments motivated me to open bug
#378241, asking courier-authlib to move its public libraries to
/usr/lib, per Policy 10.2, which would remove the necessity of
courierpassd using rpath.

Steve Langasek followed up with the bug, wisely decreasing its
severety and stating:

   I think that suggestion in policy is worse than using rpath and that
   this ought to be revised.

With that extended background, here are my two specific requests:

1) It would be really nice if policy said something about rpath.
   Currently the most official documentation about it which I could
   find is:

  http://wiki.debian.org/RpathIssue

   If rpath is going to be checked for by lintian, then it seems
   reasonable to me that there be some official documentation
   explaining Debian's position on the matter. I've currently asked
   for this in lintian's bug #378054, but perhaps there should be a
   parallel bug in debian-policy or elsewhere?

2) It would be nice to have some further discussion on Steve's comment
   regarding the relative merit of using rpath and /usr/lib/package
   versus moving all shared libraries into /usr/lib. I opened bug
   #378055 in lintian about this, but given the developments which
   have taken place since then, it is no longer clear to me what the
   right thing is for lintian to do. Again, I think that
   clarification on this is in policy's domain.

thanks,
Charles

-- 
A beard
That's rough
And overgrown
is better than
A chaperone
Burma-Shave
http://burma-shave.org/jingles/1951/a_beard


signature.asc
Description: Digital signature


Re: rpath and /usr/lib/package directories

2006-07-15 Thread Goswin von Brederlow
Charles Fry [EMAIL PROTECTED] writes:

 Hi,

 I recently uploaded courierpassd, which wraps some functionality of
 courier-authlib in the popassd protocol. lintian has been warning me
 about the use of rpath, about which I posted on debian-mentors. This
 culminated in the following thread:

 http://lists.debian.org/debian-mentors/2006/07/msg00221.html

 Which along with several other related comments motivated me to open bug
 #378241, asking courier-authlib to move its public libraries to
 /usr/lib, per Policy 10.2, which would remove the necessity of
 courierpassd using rpath.

 Steve Langasek followed up with the bug, wisely decreasing its
 severety and stating:

I think that suggestion in policy is worse than using rpath and that
this ought to be revised.

 With that extended background, here are my two specific requests:

 1) It would be really nice if policy said something about rpath.
Currently the most official documentation about it which I could
find is:

   http://wiki.debian.org/RpathIssue

If rpath is going to be checked for by lintian, then it seems
reasonable to me that there be some official documentation
explaining Debian's position on the matter. I've currently asked
for this in lintian's bug #378054, but perhaps there should be a
parallel bug in debian-policy or elsewhere?

 2) It would be nice to have some further discussion on Steve's comment
regarding the relative merit of using rpath and /usr/lib/package
versus moving all shared libraries into /usr/lib. I opened bug
#378055 in lintian about this, but given the developments which
have taken place since then, it is no longer clear to me what the
right thing is for lintian to do. Again, I think that
clarification on this is in policy's domain.

 thanks,
 Charles

Please note that this (rpath) prevents automatic multiarch conversion
for packages. Instead of a simple post procesing of the deb files a
much more complicated change has to be made to change the rpath. It
also requires the use of extra -L statements during build.

In conclusion /usr/lib/package for shared libraries just makes
everybodies live more complicated.

MfG
Goswin

PS: This would affects amd64-libs, ia32-libs and OOo for amd64 if any
component uses the /usr/lib/package way.


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: rpath and /usr/lib/package directories

2006-07-15 Thread Steve Langasek
On Sat, Jul 15, 2006 at 07:50:42AM -0400, Charles Fry wrote:
 2) It would be nice to have some further discussion on Steve's comment
regarding the relative merit of using rpath and /usr/lib/package
versus moving all shared libraries into /usr/lib.

Uh, that was *not* the recommendation that you were citing from policy, nor
was it the one I was suggesting should be overturned.

I was responding to the suggestion that rpath was inappropriate, but adding
your own directory to /etc/ld.so.conf would be ok.  It would not be ok;
there *is* no compelling reason for a package to add its own subdirectories
to /etc/ld.so.conf.  If the libraries are for public consumption, they
should be installed to /usr/lib like all the others and share the simple,
flat namespace cleanly; if they are not for public consumption, they should
be kept in their own subdirectory and a mechanism such as rpath,
LD_LIBRARY_PATH in a wrapper script, or dlopen()ing used instead.

-- 
Steve Langasek   Give me a lever long enough and a Free OS
Debian Developer   to set it on, and I can move the world.
[EMAIL PROTECTED]   http://www.debian.org/


signature.asc
Description: Digital signature


Re: rpath and /usr/lib/package directories

2006-07-15 Thread Steve Langasek
On Sat, Jul 15, 2006 at 02:25:49PM +0200, Goswin von Brederlow wrote:
 Please note that this (rpath) prevents automatic multiarch conversion
 for packages. Instead of a simple post procesing of the deb files a
 much more complicated change has to be made to change the rpath. It
 also requires the use of extra -L statements during build.

 In conclusion /usr/lib/package for shared libraries just makes
 everybodies live more complicated.

Which is no different from all the other mechanisms used for lookups of
private modules; they all require source modification for paths to be fixed
to allow co-existence in a multiarch environment.  The only way through this
is to define multiarch-clean paths and encourage their use.

-- 
Steve Langasek   Give me a lever long enough and a Free OS
Debian Developer   to set it on, and I can move the world.
[EMAIL PROTECTED]   http://www.debian.org/


signature.asc
Description: Digital signature


Bug#378386: [PROPOSAL] Include the GFDL in the set shipped in /usr/share/common-licenses

2006-07-15 Thread Adeodato Simó
Package: debian-policy
Version: 3.7.2.1
Severity: wishlist
Tags: patch

Hello.

In [1], Joerg Jaspert initiated discussion about including more licenses
in the set of licenses shipped in /usr/share/common-licensed. As part of
the discussion I inquired in [2] about the possibility of adding the GFDL
to that set, given the number of packages that ship some contents under
this license.

Since at least one of the Policy editors, namely Manoj Srivastava,
agreed in [3] that the GFDL passes the criteria for inclusion, I'm now
transforming the mentioned informal discussion into a formal request.

  [1] http://lists.debian.org/debian-policy/2006/06/msg00100.html
  [2] http://lists.debian.org/debian-policy/2006/06/msg00109.html
  [3] http://lists.debian.org/debian-policy/2006/06/msg00206.html

Attached is a patch with the possible wording for the change.

Also, if somebody wonders as for why the change is not first commited in
the base-files package, and let Policy catch up on reality later: as per
/usr/share/doc/base-files/FAQ, the base-files has delegated that decision
to the Policy editors.

I will submit a bug against base-files after an updated debian-policy is
uploaded, or this bug marked by pending by one of the editors.

Thanks!

-- 
Adeodato Simó dato at net.com.org.es
Debian Developer  adeodato at debian.org
 
We learned that the Linux load average rolls over at 1024. And we
actually found this out empirically.
-- H. Peter Anvin from kernel.org
--- ./policy.sgml~  2006-07-16 00:14:47.0 +0200
+++ ./policy.sgml   2006-07-16 00:17:56.0 +0200
@@ -8618,8 +8618,8 @@
 
p
  Packages distributed under the UCB BSD license, the Artistic
- license, the GNU GPL, and the GNU LGPL should refer to the
- corresponding files
+ license, the GNU GPL, and the GNU LGPL, or ship documentation
+ under the GFDL, should refer to the corresponding files
  underfile/usr/share/common-licenses/file,footnote
 p
   For example, 
@@ -8627,6 +8627,7 @@
   file/usr/share/common-licenses/BSD/file,
   file/usr/share/common-licenses/GPL/file,
   file/usr/share/common-licenses/LGPL/file,
+  file/usr/share/common-licenses/GFDL/file,
   file/usr/share/common-licenses/GPL-2/file, and
   file/usr/share/common-licenses/LGPL-2.1/file, and so on.
 /p


Re: rpath and /usr/lib/package directories

2006-07-15 Thread Goswin von Brederlow
Steve Langasek [EMAIL PROTECTED] writes:

 On Sat, Jul 15, 2006 at 02:25:49PM +0200, Goswin von Brederlow wrote:
 Please note that this (rpath) prevents automatic multiarch conversion
 for packages. Instead of a simple post procesing of the deb files a
 much more complicated change has to be made to change the rpath. It
 also requires the use of extra -L statements during build.

 In conclusion /usr/lib/package for shared libraries just makes
 everybodies live more complicated.

 Which is no different from all the other mechanisms used for lookups of
 private modules; they all require source modification for paths to be fixed
 to allow co-existence in a multiarch environment.  The only way through this
 is to define multiarch-clean paths and encourage their use.

I think the case in discussion is not about private modules. I agree
that for private modules one is screwed no matter what existing
solution is used. Private modules (or plugins) often are listed in a
conffile which adds yet another problem.

Multi-arch paths for both conffile and plugins is indeed the only sane
solution there.

MfG
Goswin


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]