Re: AdaCurses: take the opportunity of the last GPR tools for build and install.

[Thomas Dickey]

On Sun, Aug 18, 2019 at 09:11:08PM -0400, Thomas Dickey wrote:
> On Sun, Aug 18, 2019 at 10:05:22AM +0200, Pascal wrote:
> > Hello Thomas,
> > 
> > > Le 17 août 2019 à 22:27, Thomas Dickey <address@hidden> a écrit :
> > > 
> > > On Sat, Aug 17, 2019 at 07:34:19PM +0200, Pascal wrote:
> > >> Hello,
> > >> 
> > >> I took the opportunity of the last GPR tools for build and install 
> > >> AdaCurses from Ada compiler suite GNAT Community 2019.
> > >> https://www.adacore.com/download
> > >> 
> > >> I took ncurses source codes from Github repo:
> > >> https://github.com/mirror/ncurses
> > > 
> > > https://invisible-island.net/personal/git-exports.html
> > > https://github.com/ThomasDickey/ncurses-snapshots
> > > 
> > >> I push all changes to my Github:
> > >> https://github.com/Blady-Com/ncurses/tree/proposal_01
> > >> 
> > >> Do I send a pull request to the ncurses Github repo?
> > > 
> > > no.
> > 
> > Ok I understand the way you proceed.
> > 
> > > 
> > >> Or else what is the correct way to submit chage proposals?
> > > 
> > > diff's, like the attached
> > > 
> > > I don't know about the change itself, since it has to be tested against 
> > > the
> > > existing gnat configurations.
> > 
> > What is your testing process?
> 
> I have scripts which do configure/make/make-install/make-uninstall,
> as well as making packages (for the AdaCurses tree by itself).
> I keep the logs, compare and resolve differences.  What platforms I build
> and test depends on what's changed recently.
> 
> Most of the builds I do are with the whole ncurses tree, which
> makes some difference in the way the configure/make process works.
> 
> But I test-build AdaCurses by itself occasionally (I have logs
> for Debian, Fedora, and OpenSUSE from last winter, for _that_).
> 
> Besides those, my whole-tree builds include cygwin and CentOS.
> 
> > On my side, I tested the changes with GNAT Community 2019 on macOS 10.13.
> 
> I did some test-builds with the 2018 version on macOS 10.13,
> which worked (though I noticed a problem with one of the pathnames).

I took a quick look at this, but it doesn't improve my ncurses6 build
for MacOS GNAT-2019 (that fails to find the libpanelw shared library
with/without the patch), and causes the Debian 9/testing builds to fail
to find _traces

(it's on my to-do list to investigate further)

The MacOS problem is one I knew about - simply haven't gotten to:

181229
**      The MacOSX gnat package insists on being installed under ~thomas/opt,
                /Users/thomas/opt/GNAT/2018/bin
        and does run when the $PATH is updated.  A test-build using cfg-ncurses6
        fails since gnat's looking for "-lpanelw", but I'm building "-lpanelw6".
        Looks like that comes from a pragma in an m4 file "Linker_Options".
        There's 4 files to fixup, probably best to generate a script which
        can be used to replace the DEL_ADAMODE hack.

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

signature.asc
Description: PGP signature