Re: [Chicken-hackers] Should we make a 4.12 release?

[Mario Domenech Goulart]

Hi Peter,

Thanks for the summary on tickets for the new release.  Please, see some
comments below about #1298.

On Sat, 31 Dec 2016 14:11:38 +0100 Peter Bex <address@hidden> wrote:

> I'm wondering whether we should make a 4.12 release candidate.  There
> are a handful of very bad bugs (#1317, #1337, #1336) and 3 CVEs that
> have been fixed.
>
> There are still 3 patches waiting (1 bugfix for #1144), so I'd prefer
> to wait until after they've been applied.
>
> After my patch for #1144 is applied, there will be 4 open bugs left:
> https://bugs.call-cc.org/query?status=assigned&status=new&status=accepted&status=reopened&group=status&milestone=4.12.0
>
> I'm not sure what to do about #1293.  We don't have a simple way
> of assigning unique object IDs, so some objects will "notice" they
> get moved by the GC and all objects will notice when they're mutated.
>
> I proposed a patch for #1294, but Felix didn't like it.  It's a tricky
> thing to fix otherwise, so Felix, unless you know how to fix this, we
> probably are better off postponing it.
>
> Mario, what is the final word on #1298?  Your last comment seems to
> indicate that this should maybe be closed "wontfix"?

I don't really know what to do about that ticket.  I don't know which
behavior is correct, 4.10's or 4.11's.

If we keep the current behavior (i.e., 4.11's, which installs the
dependencies in the argument for -prefix), packagers' life is going to
be more difficult, as they need to determine which files an egg
installs.  As far as I can see, the only way to do that is extracting
file paths out of the `files' form in .setup-info files.  But to do
that, packagers first need to determine which .setup-info files eggs
install, as a single egg may install multiple .setup-files.  To do that,
they have to scan all .setup-info files in the local egg repo directory
and filter the ones whose `egg' form match the egg being packaged.  Only
after determining the right .setup-info files they can find which files
to package.  This approach is not perfect, though, as .setup files can
install files into the local egg repo using methods other than
install-{program,script,extension} (then the installed files won't go to
.setup-info's `files' form).

One solution would be adding another option to chicken-install (e.g.,
-no-install-deps, see [1]).  With this option, we could have both
behaviors (i.e., 4.10's and 4.11's).  But then it's one more relatively
obscure option for chicken-install.

What do people think?

Felix: what was the original intended behavior for -prefix, regarding
handling of dependencies?

[1] http://parenteses.org/mario/misc/no-install-deps.patch

> Christian was unable to check whether #1107 has been fixed with the
> latest GCC, because the port was broken.  It's just a warning which
> has probably been in there forever.  We could add the --std switch,
> but I have no idea about the consequences, which could be pretty
> far-reaching for eggs (or could make no difference at all).

All the best.
Mario
-- 
http://parenteses.org/mario