Re: Cross Toolchains "Squeeze" Release Goals
Selon Hector Oron <email@example.com>:
> 2009/8/12 Ludovic Brenta <firstname.lastname@example.org>:
>> What kind of target do you have in mind? armel bare-board? with uClibc?
>> with EmDebian? with full Debian?
> I personally do not know a bit on ada, I should learn sometime, as I
> have heard it is very suitable for satefy systems. The target really
> does not matter as what we try, IIRC, it is already supported by
> upstream, we just need gcc-source package to produce an
> arm-linux-gnueabi-gnat compiler. The binaries will be part of Emdebian
> repository and if Debian agrees and it has a doable amount of users,
> Debian might be interested on having those. But, surely you should be
> able to build a cross gnat package in from Debian sources when this
> work is done.
Yes, the target does matter; we need not only a cross-compiler but also a
run-time library (libgnat-4.4.so.1 and/or libgnat.a). The run-time library
is needed for exceptions and for multitasking (Ada has multitasking built
in). On GNU/Linux, the compiler translates tasking constructs into calls to
POSIX Threads (i.e. glibc or eglibc), so the Ada run-time library requires it
along with a Linux or FreeBSD kernel. OTOH, on bare-board targets, the
compiler and run-time library can together implement a minimalistic
multitasking kernel. There are additional targets running various embedded
real-time kernels but I don't think we'll have the manpower to test those.
> If you were asking me for a test platform, I have one balloonboard and
> a sheevaplug which we can test on.
It is a bit premature right now for me to ask a test platform; I'd like to
first understand our goals and strategy in more details than a nice-sounding
but vague "cross-compiler".