[Bug 478920] Review Request: globus-xio-popen-driver - Globus Toolkit - Globus XIO Pipe Open Driver

[bugzilla at redhat.com]

Please do not reply directly to this email. All additional
comments should be made in the comments box of this bug.


https://bugzilla.redhat.com/show_bug.cgi?id=478920


Mattias Ellert <mattias.ellert at fysast.uu.se> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|Review Request:             |Review Request:
                   |globus-xio-popen-driver -   |globus-xio-popen-driver -
                   |Globus Toolkit - Globus XIO |Globus Toolkit - Globus XIO
                   |BW Limit Driver             |Pipe Open Driver




--- Comment #4 from Mattias Ellert <mattias.ellert at fysast.uu.se>  2009-04-24 07:51:10 EDT ---
(In reply to comment #3)
> I have a couple questions for this one:
> 
> * Please don't abbreviate BW in the description. Is it bandwidth?

The package description was kind of weird, wasn't it? It didn't really fit the
name of the package. My first response to your question was "of course it means
bandwidth", but since I really had no backup for such a claim I started to look
at the sources to figure it out. This package has no doxygen markup and not
much other comments either, so this was not easy.

I ended up writing an e-mail to the upstream author of this package asking him
what the package description meant. It turned out that in this case the package
description was way off, as it was the same as in the XIO skeleton driver
template package (source-trees/xio/drivers/skeleton/source) and hadn't been
adopted to describe what this driver is doing, when the skeleton had been used
as a basis for this package.

The upstream author has now fixed the description in the upstream CVS:

http://viewcvs.globus.org/viewcvs.cgi/xio/drivers/popen/source/pkgdata/pkg_data_src.gpt.in?r1=1.3&r2=1.4&view=patch

I have added this change as a patch to the updated package. I also added an
additional sentence to the long description, since also the new description is
a bit cryptic.

> ? Do we really need to use the word "Globus" twice in the summary for this
> package?

Since this is in the autogenerated part of the specfile, I'd like to keep it.

> ? # This library is opened using lt_dlopenext,...
> Uhm, why? How many packages do that? Can't those packages be "fixed" instead?

The XIO framework has a number of built-in drivers. But it is extendible by
using plugins that contain additional drivers. These plugins are based on the
libtool dynamic loader library (libltdl).

The libltdl, when opening a plugin, reads the .la file to find out the real
library name .so.X and then opens this library. If the .la file is not there it
can not find which library to load. So runtime access to the .la file is
needed. There is really no way to "fix" this as long as upstream uses libltdl
to load plugins.

I documented this feature in the Plugins section of the Packaging Draft:
http://fedoraproject.org/wiki/PackagingDrafts/Globus#Plugins

Of the 30 globus packages I have submitted for review 2 are plugins that are
opened using this mechanism.

Updated package with corrected package description:
http://www.grid.tsl.uu.se/repos/globus/info/new/globus-xio-popen-driver-0.2-2.fc10.src.rpm
http://www.grid.tsl.uu.se/repos/globus/info/new/globus-xio-popen-driver.spec

-- 
Configure bugmail: https://bugzilla.redhat.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.