svn commit: r369236 - in head/science: ghemical libghemical libint mpqc

[John Marino]

On 9/25/2014 14:41, Tijl Coosemans wrote:
> On Thu, 25 Sep 2014 13:37:01 +0200 John Marino <freebsd.contact at marino.st> wrote:
>> The previous commit changes all the extensions from 8.0.1 to 7.1.0.
>> This puts it back to how it was before.  The libtool caused a regression
>> (it's not the first time I've run into this) so this commit fixes that
>> regression manually.
> 
> The extension was .so.8 and the previous commit changed it into
> .so.7.1.0.  That is not a regression.  

More accurately it changed it from .so.8 to .so.7 and .so.7.1.0

>It is supposed to work like
> that.  That's how libtool works on Linux, Darwin and a number of
> other systems and USES=libtool makes it work like that on FreeBSD
> as well.
> 
> The advantage of this new version scheme is that the major number
> doesn't change as much.  In the new scheme these libraries are still
> at version 7 and have not been bumped to 8 yet like in the old scheme.
> This is a good thing.  Fewer changes to the major version means fewer
> cases where dependent ports need a PORTREVISION bump.

So it would appear that when one selects USES=libtool, shared library
version changes are common, expected, and according to above, desired.
I have to admit that I was surprised to see a replacement tool change
the version numbers; I didn't expect it.

Did your libtool writeup on email make it into a wiki?  and is this
effect of switching from USE_LIBTOOL to USES=libtool noted anywhere?

I was actually trying to ping you on IRC about this last night.  I ended
up wasting a lot of time on something that's going to get reverted again. :(

John