> Currently we simply pass the '-revision 3.0' option to libtool and
> it decides where to stick it. I personally don't see much difference
> between naming the library libxerces-c-3.0.so or libxerces-c.so.3.0.
> Either way the version number is embedded into the name.

There's a big difference. On Unix, the linker is designed to bind the
executable to a "neutral" instance of the library that allows for minor
upgrades without rebuild. The actual build command is supposed to find
libfoo.so, but it binds libfoo.so.major into the file.

The actual library itself is libfoo.so.major.rev so that at runtime you get
the latest rev of libfoo.so.major.

What you did in the past worked more or less because the order of the
extensions was consistent with this, even if it didn't exactly work the same
way. The current makefile is just flat broken. It's a bug, not a choice.

> I did some more thinking about -version-info vs -revision. I think
> I now understand things better and we can do it either way. With
> -version-info, the version passed will be different from the release
> version.

That is by design, they are not the same concept.

> Because Xerces-C++ release numbers are governed by the
> interface compatibility (e.g., major releases are backwards-
> incompatible, minor releases are interface-compatible but not binary
> compatible, and build releases are binary compatible), the age
> component will always be 0.

That's fine, that's simply a product mgmt choice.

> As a result, I don't see much benefit
> from using a more complex -version-info over the simpler -release.

One is wrong, one is right. It's simply that direct at the moment. It may be
that you can get -release to give you output consistent with the 2.x build,
but right now it's not.

> Another drawback of using -version-info is that we will have to
> somehow replicate this in the ICU message id which currently
> looks like this: xercesc_messages_3_0.

Why is it a requirement for them to be connected?

-- Scott



---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to