Re: Fwd: Version of _nc_write_object symbol

[Thomas Dickey]

On Mon, Jan 18, 2016 at 03:32:38PM +0100, Miroslav Lichvar wrote:
> On Sun, Oct 11, 2015 at 01:01:01PM -0400, Thomas Dickey wrote:
> > | The 20150905 patchlevel introduced a new symbol _nc_write_object
> > | tagged
> > | with the version NCURSES_TIC_6.0.current.  I assume that's not the final
> > | version, but is it same to assume that it won't change before a 6.1 
> > prerelease?
> > 
> > yes - I'm uncertain of the best way to evolve this, but in this case I did 
> > the
> > same process as I did for the 5.9/6.0 development, by marking new symbols as
> > "current", and when I got to the end of 5.9 changes, changing "current" to a
> > specific patch-date.
> 
> Would it make sense to use the patch date right when the symbol is
> introduced? In distributions that package patch snapshots of ncurses
> the loss of the original symbol version would require rebuild of
> client applications that use it. Or is that an intention?

yes... I debated (with myself) on this detail, thinking that

        a) I'd prefer tracking regular releases on the symbol versioning
           in part, because I generated the files by compiling and
           extracting the information -- and limiting the number of
           datapoints keeps it manageable.

        b) development snapshots are just that - development.  For releases
           I do a lot more test-builds to see that everything that I can
           build on still builds/works.  Usually there are fixes needed,
           but the end result should be more stable than the development
           snapshots.

I realize it's a nuisance to rebuild things for a regular release, but
thought that was an acceptable cost for the (presumably) more stable
result.

-- 
Thomas E. Dickey <address@hidden>
http://invisible-island.net
ftp://invisible-island.net

signature.asc
Description: Digital signature