MobileRead Forums - View Single Post - Qt for Embedded Linux, minimum requirements not met, there is always a way

knc1 · 09-04-2012, 08:57 AM

@NiLuJe : I don't recall posting that there was a problem with ct-ng.

There is a "problem" with the C++ code of one (or two) of the libraries used for the "graphite" optimizations - like one of them has never been ported to any machine other than x86 & x86_64.

Not a problem when only building a cross-compiler (since the optimization only has to run on the host).
But becomes a problem when doing a full three stage 'build and test' of gcc for non-Intel.
So our new, native compiler may have that optimization method disabled (unless I can port the code to arm).

Since ct-ng only does a bootstrap + stage one build, it never has that problem.
Yann E. Morin still has doing a full 'build and test' on his to-do list.
And he told me that he does not expect to have the time to do that work this year.

But he will accept 'patches' for that work (if I can figure out how to do it).

The same story with the hidden 'build native' on the configuration menu (no working code behind it, so its hidden) and no time to work on it this year.
But he will accept 'patches' if we want to code it for him.

Translation of all above:
Somebody has to 'engineer' this solution at least once by hand before it can be automated and YEM doesn't have time to help us.

09-04-2012, 08:57 AM	#13
knc1 Going Viral Posts: 17,212 Karma: 18210809 Join Date: Feb 2012 Location: Central Texas Device: No K1, PW2, KV, KOA	@NiLuJe : I don't recall posting that there was a problem with ct-ng. There is a "problem" with the C++ code of one (or two) of the libraries used for the "graphite" optimizations - like one of them has never been ported to any machine other than x86 & x86_64. Not a problem when only building a cross-compiler (since the optimization only has to run on the host). But becomes a problem when doing a full three stage 'build and test' of gcc for non-Intel. So our new, native compiler may have that optimization method disabled (unless I can port the code to arm). Since ct-ng only does a bootstrap + stage one build, it never has that problem. Yann E. Morin still has doing a full 'build and test' on his to-do list. And he told me that he does not expect to have the time to do that work this year. But he will accept 'patches' for that work (if I can figure out how to do it). The same story with the hidden 'build native' on the configuration menu (no working code behind it, so its hidden) and no time to work on it this year. But he will accept 'patches' if we want to code it for him. Translation of all above: Somebody has to 'engineer' this solution at least once by hand before it can be automated and YEM doesn't have time to help us. Last edited by knc1; 09-04-2012 at 09:15 AM.