On 08/17/2010 01:38 AM, Sune Vuorela wrote:
> On Tuesday 17 August 2010 01:04:55 Steffen Möller wrote:
>> Would it be reasonable to possibly perform this second build as an
>> extension of your regular qt4-x11 package (which would then somehow have a
>> wrong name, but who cares) with the --embedded flag set. What comes to
> I have thought of this more than once, but it has mostly stalled for various
> reasons (beside the time involved in doing it)
divide that time by 3/2 (some overhead for parallelism)
> 1) qt4-x11 source package is already a buildd killer (both in time and space
> usage, 3-4 days of build time, more than 4G of space)
I agree. And one should probably try to find ways to reduce those
(still) rather enormous resources. On the other hand, this is exactly
the benefit that the Debian packaging would bring....plus some more.
> 2) for most real usages of Qt/linuxfb, you would want a specific feature set
> that you actually need, and then need to recompile anyways
they have some plugin mechanism that the Debian packages could possibly
mimic in some way. The first packaging will not be perfect. To me such
optimisations are likely to go eventually into the production system. I
am more after allowing development more easily.
> and the most important one
> 3) qt/linuxfb is not binary compatible with qt/x11, so for all relevant
> libraries and applications, we would also need to do two build passes, one
> the linux stuff and one for x11, making it not really that interesting.
As of today, there is no alternative to using X on your desktop, and there
probably won't be any such alternative any time soon. This is not what
I am after.
But for embedded devices, you often have just a single application to be
run. For the very very deeply embedded deviced you would not even use
Qt, but for the upcoming microwaves or washing machines or fridges or
heating devices I could well imagine some basic Qt application to run
and while developing that you don't care too much if the Qt library is
perfectly optimised for your platform ... I hope. And even then, to have
a Debian package to start from for your builds is probably helpful.
> The api is the same, so you can still, in 99.5 or so of the cases do the
> development against qt/x11 and just deploy it against qt/Linuxfb
But that deployment is not within the reach of Debian, not even for the
development environments. If the qvfb helps to get from those 99.5% up
to 99.95%, then this should already be rather valuable for the
developers' peace of mind. And without that deployment we don't see the
discussions towards optimisations or general abstractions of the boot
process and other bits.
I am still in favour of it. Maybe we can help the buildds in some way.