Re: lib/42182: libpixman-1.so.0 is no longer available in NetBSD 5.0_STABLE

["David H. Gutteridge" <dhgutteridge%sympatico.ca@localhost>]

The following reply was made to PR lib/42182; it has been noted by GNATS.

From: "David H. Gutteridge" <dhgutteridge%sympatico.ca@localhost>
To: Soren Jacobsen <snj%pobox.com@localhost>
Cc: gnats-bugs%NetBSD.org@localhost
Subject: Re: lib/42182: libpixman-1.so.0 is no longer available in NetBSD 
5.0_STABLE
Date: Sat, 31 Oct 2009 19:28:01 -0400

 On 2009-10-31, at 5:50 PM, Soren Jacobsen wrote:
 
 > On Oct 31, 2009, at 11:45 AM, David Gutteridge wrote:
 >
 >> The following reply was made to PR lib/42182; it has been noted by  
 >> GNATS.
 >>
 >> From: David Gutteridge <dhgutteridge%sympatico.ca@localhost>
 >> To: gnats-bugs%netbsd.org@localhost
 >> Cc:
 >> Subject: Re: lib/42182: libpixman-1.so.0 is no longer available in  
 >> NetBSD 5.0_STABLE
 >> Date: Sat, 31 Oct 2009 14:08:30 -0400
 >>
 >> Those two libraries are also not covered off by the postinstall(8)
 >> obsolete check.  So an upgrade masks the problem.  (I've just
 >> encountered this.)
 >
 > What do you mean covered off by the postinstall obsolete check?
 >
 > Removing the old libraries would make the situation even worse, and  
 > I wouldn't call leaving them in place masking the problem.
 >
 > I'm not sure there's really much that can be done about this PR.   
 > We'll need to have separate 5.0 and 5.1 pkg binaries, but after an  
 > upgrade, no previously installed packages should break, because the  
 > old libraries will still be installed.
 >
 > pkg_add has always (or at least for quite a long time -- I can't say  
 > off of the top of my head whether this check has been there since  
 > the very beginning) complained about version differences, and this  
 > is one of the cases "protected" (in the sense of "yes, I admit it  
 > sucks, but hey, technically we did warn you") by this check.
 >
 
 I must have just misunderstood the meaning of "obsolete", then.  I took
 it to be functionality that shows any differences from what the state
 of the system would be if a clean install had been done instead of an
 upgrade from source.  (Local customizations aside.)  But it's more
 nuanced/sophisticated/specialized than that, then.  (My notion of it in
 part was as a support tool to explain differences between systems with
 the same release branch and CVS source date.  Obviously that's not its
 primary purpose, regardless.)
 
 By "masks the problem", I meant the version mismatch probably hasn't
 turned up for too many people, so that's why there wasn't much traffic
 about this topic.  I was alluding more to the fact I hadn't seen it
 documented elsewhere, and was offering an explanation as to why it
 might've been missed.  (Here's where someone will show me an incredibly
 obvious place it was mentioned that I overlooked...)
 
 Sorry for the misunderstanding,
 
 Dave