[argyllcms] Re: Argyllcms 1.0.1 packaged in fedora-devel

On Mon, Jul 28, 2008 at 3:58 AM, Graeme Gill <graeme@xxxxxxxxxxxxx> wrote:
> Frederic Crozat wrote:
>
>> I do. But you should also respect people who want to improve the
>> entire platform.
>
> And I'm not preventing you in any way. Since Argyll doesn't
> use the system libusb, you can do anything you want to it.
>
> By all means encourage me to get my changes adopted up
> stream, so that system libraries are improved, and Argyll
> can just use them, but don't try and force my hand on it.
> Making sure the application works as intended is far more
> important to the end users than general "goodness" such as
> improving various platforms.

As I wrote in another mail, I understand you are more interested in
working on Argyllcms
than improving the stack.

Maybe now is the right time to start testing the new libusb-1.0
rewrite to ensure this first
version will be working as expected for Argyllcms, when it hits its
first stable release.

>> And again, I ask the question : where did I broke Argyllcms 1.0.0 on
>> Mandriva Linux Cooker ?
>
> I have no idea whether it's broken or not, and that's the core point.
> It's a QA (Quality Assurance) issue. As soon as you make a change such
> as you have (switching from the provided library to one that by default
> is known to have problems with this particular application), you invalidate
> the QA testing done on the configuration you started with. Perhaps
> you have patched the libusb in your system appropriately (good for
> you if you have!), but how are you now testing the application to
> make sure that it works as well as what you started with it, given
> that you know the change might have broken it, and you patch not
> have fixed it ?

Well, that is the point of development distributions and wide tester community.
We don't have all possible hardware available so we rely on community
to test stuff and
report regressions. And on average, it works quite well (it spotted an
issue with your initial
libusb patch). And when I say 'all possible hardware", I'm not only
speaking about Argyllcms but
all USB devices (since the patch is about libusb).

And since I use your exact patch on our libusb, I have to trust you
about it, nobody else ;)

> Unless you have a solution to this (and why you'd want to take
> on the QA of an application such as Argyll that requires particular
> equipment and manual testing is a mystery to me), I really don't
> think you should be doing it. And no, sorry, I've got enough on my
> plate without trying to do QA on versions of Argyll that others
> have taken it upon themselves to modify.

Again, you are focusing on your specific development, I'm not. I want
to improve the plaform : pushing
argyllcms in the distro is improving the platform, pushing your libusb
patch is improving the platform. Both are
linked, but not that strongly IMO, because your patch was not needed
for all kind of probes (my own Huey was working
fine without it) and I didn't want to push hacks in the platform just
to fix one probe (which is why I didn't integrate your
initial libusb patch).

Anyway, I think we have rehashed the problem enough now. Let's see if
we can get libusb "fixed" in distributions out there ;)

-- 
Frederic Crozat

Other related posts: