Re: Possible library versioning approach -- (evaluation requested)

[Rob Browning <rlb@defaultvalue.org>]

Bill Allombert <allomber@math.u-bordeaux.fr> writes:

> You do not need to use a table:
> you can use 'explicit' soname:
>     glibversion  guileversion  gwrap-glib-1-soname
>         1.2          1.4       1-glib1.2-guile1.4
>         1.2          1.6       1-glib1.2-guile1.6
>         2.0          1.4       1-glib2.0-guile1.4
>         2.0          1.6       1-glib2.0-guile1.6
>
> I am not sure this is the right solution, but this scale much better
> than using 1,2,3,4.  and it will be easier to handle in the long
> run, and is self-explanatory. Also the soname is not mixed with the
> name.

Thanks, I'd actually thought about that as a general solution but
dropped it.  It seemed like if libs always did this, and sorted the
soname-lib-names alphabetically, then the more general binary
compatibility problem would be solved.  I quit thinkin about it
because I just presumed that this approach would be too unconventional
to have much chance at wide adoption.  I also didn't know how well
such an approach would work cross-platform -- whether or not sonames
could be arbitrary on all of the important platforms.  More
immediately, I didn't think libtool (which is what a large number of
packages, including gwrap use these days) would allow you to directly
specify a soname.  Although they're admittedly more obscure, plain
integers seemed more likely to side-step the above issues.

-- 
Rob Browning
rlb @defaultvalue.org, @linuxdevel.com, and @debian.org
Previously @cs.utexas.edu
GPG starting 2002-11-03 = 14DD 432F AE39 534D B592  F9A0 25C8 D377 8C7E 73A4