Bug#354115: bug fix causes problems

2006-08-28 Thread Santiago Vila
On Sat, 26 Aug 2006, Andreas J. Guelzow wrote:

 To drop the 
 order bind, hosts
 line from /etc/host.conf is very ill advised. 
 
 It just took me a large number of hours to track the problem that an
 essential (for us) software package stopped working thanks for this
 unnecessary bug fix.  I call it unnecessary since that line is simply
 not used in recent versions of glibc.
 
 This program that failed for us is Mathematica version 4.0. This is a
 statically linked program that uses the order line of /etc/host.conf.
 (While some in our department have migrated to maxima, several members
 are still using Mathematica, and failure to have this run correctly is a
 major problem!)
 
 Please don't drop such items with out regard to older pieces of
 software!

The reason to remove the line is not that it's no longer used,
the reason is that it's *misleading*.

Now you have a point, some software may break because of this,
but so far the only example happens to be jurassic statically linked
software which is also non-free and proprietary (and of course
Debian does not distribute). I don't feel a lot sorry for this.

I think this is the kind of problem we should better document, instead
of perpetuating the misleading line. May I suggest that you report
this against the release notes document for etch?


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



Bug#354115: bug fix causes problems

2006-08-28 Thread Andreas J. Guelzow
On Mon, 2006-28-08 at 19:25 +0200, Santiago Vila wrote:
 The reason to remove the line is not that it's no longer used,
 the reason is that it's *misleading*.

That could easily be fixed by adding a comment line that the order
line is only used by old versions of the libc library. There is no need
to break old programs.

 
 Now you have a point, some software may break because of this,
 but so far the only example happens to be jurassic statically linked
 software which is also non-free and proprietary (and of course
 Debian does not distribute). I don't feel a lot sorry for this.

This issue has nothing to do with the fact that Mathematica is non-free
and proprietary. The problem would occur with any statically linked
program that happens to use an older version of libc, whether free or
not.

 
 I think this is the kind of problem we should better document, instead
 of perpetuating the misleading line. May I suggest that you report
 this against the release notes document for etch?

I am not sure how the manual page for /etc/hosts.conf is anymore
misleading now then it was before. I am quoting from the current page:

---
The  file /etc/host.conf contains configuration information specific to
   the resolver library.  It should contain one configuration
keyword  per
   line,  followed by appropriate configuration information.  The
keywords
   recognized are order, trim, multi, nospoof, spoof, and reorder.
These
   keywords are described below.

   order  This keyword specifies how host lookups are to be
performed.  It
  should be followed by one or more lookup methods,
separated  by
  commas.  Valid methods are bind, hosts, and nis.
---

I gather Debian does not care about backwards compatibility.


Andreas



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



Bug#354115: bug fix causes problems

2006-08-26 Thread Andreas J. Guelzow
To drop the 
order bind, hosts
line from /etc/host.conf is very ill advised. 

It just took me a large number of hours to track the problem that an
essential (for us) software package stopped working thanks for this
unnecessary bug fix.  I call it unnecessary since that line is simply
not used in recent versions of glibc.

This program that failed for us is Mathematica version 4.0. This is a
statically linked program that uses the order line of /etc/host.conf.
(While some in our department have migrated to maxima, several members
are still using Mathematica, and failure to have this run correctly is a
major problem!)

Please don't drop such items with out regard to older pieces of
software!

Andreas
-- 
Prof. Dr. Andreas J. Guelzow
Dept. of Mathematical  Computing Sciences
Concordia University College of Alberta



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