rpms/git-core/devel .cvsignore, 1.10, 1.11 git-core.spec, 1.9, 1.10 sources, 1.10, 1.11

[Ralf Corsepius]

On Tue, 2005-11-01 at 13:26 -0500, Warren Togami wrote:
> Chris Wright wrote:
> > * Ralf Corsepius (rc040203 at freenet.de) wrote:
> > 
> >>Here, the "BR: libz >= 1.2" is a "necessary condition" (in the
> >>mathematical sense) to force users (or the buildsystem) to use a
> >>sufficient zlib. Whether or not the sources support "graceful degrating"
> >>to other internal underneath, isn't of any importance.
> > 
> > 
> > Problem is BR zlib >= 1.2 is completely irrelevant when installing a
> > package, simply because implicit dependency on libz.so.1 is inaccurate.
> 
> Exactly.
True, but ... you are trying to fix a non-issue.

This issue can only occur if a user installs a git-core having been
compiled against libz.so.1.2.* on a system being equipped with
libz.so.1.1*.

The only situations this (theoretically) can happen is the user trying
to _manually_ install a Fedora git-core package on non-Fedora rsp.
ancient RH systems.

During automated updates/upgrades (up2date, yum, apt) this issue
shouldn't occur, because then, zlib should be updated/upgraded at the
same time.

> > Since this shouldn't be an issue for FC/FE users, I can just drop that
> > bit from upstream.  It's there only because somebody hit it, complained,
> > and sent a patch.

The problem with your current spec is it being inconsistent:

You currently have:
BuildRequires: zlib-devel
Requires: zlib >= 1.2

This is inconsistent. It allow building against zlib < 1.2 but
dissallows running it on such systems.


Either:
BuildRequires: zlib-devel >= 1.2
Requires: zlib >= 1.2

Or
BuildRequires: zlib-devel

Or even
BuildRequires: zlib-devel >= 1.2
Conflicts: zlib < 1.2

would be consistent.

Ralf