RTEMS 4.6 and tools

[Ralf Corsepius]

On Tue, 2004-01-20 at 03:40, Steven Johnson wrote:
> Thanks for the responses.
> 
> I do need to run it on thumb, so this has shwn me I need to use gcc 3.3.x to 
> reliably use thumb, and that there are a number of interested people in this area.
> 
> Im interested in helping get any existing thumb code integrated into a possible 
> rtems 4.6.1.  Thomas, a diff between your version of RTEMS and the version of 
> RTEMS you started it from would be a big (massively huge) start, I cant imagine 
> integrating those changes into the current head of the 4.6 branch would be an 
> overwhelming task.  There only seemed to be about 6 small bits of assembly to 
> option for arm, so the job on that side doesn't seem very big.  I agree that so 
> close (days away :) to release, none of this should affect 4.6.0.
> 
> I can build the various gcc versions myself no problems, my questions for these 
> however are:
> 
> 1.  Apart from the "Heaps of warnings" do these "newer" version of gcc cause any 
> known problems with the rtems kernel?
Hard to tell because probably nobody has tested a gcc-3.3.x compiled
RTEMS on actual hardware yet.

Some of the compile warnings I see, however give reasons to be
concerned.

Another issue with gcc-3.3.x is Ada. Ada in gcc-3.3 can't be
cross-compiled at all and therefore is unusable for RTEMS.

Nevertheless, I'd recommend to use at least gcc-3.3 for gcc development,
because gcc-3.4 is to be released and actual gcc-development has
switched to gcc-3.5.

>   Compile warnings I can live with (maybe 
> even remove).

> 2.  Are there patches for gcc 3.3.x for rtems,
Yes, in particular some relevant to you (arm-rtems).

>  or is vanila gcc ok as is?
No.

>   If there are patches, where do I get them?
Good question. They already are several months old, but I don't know if
OAR has put them on ftp anywhere - I guess not.

> 3. If I use the RTEMS CVS site for this arm/thumb further development (to track 
> a potential 4.6.1 or 4.7), should I use the rtems-4-6-branch for development
The rtems-4-6-branch is "current official stable RTEMS" and in general
should only marginally differ from the "official stable RTEMS" tarballs.

>  on this, or should I use another branch?

The "other branch" is CVS-HEAD. The plan is to fork a rtems-4-7-branch
in not too distant future, esp. once gcc-3.3.3, gcc-3.4 and rtems-4.6
have been released.

At present time, the actual RTEMS code in CVS-HEAD has not diverged
substantially from RTEMS-4.6, but its configuration and Makefiles are 
completely different from what is in RTEMS-4.7. Further radical changes
are not unlikely to happen, nevertheless, RTEMS-CVS-HEAD should be in a
usable and buildable shape, but YMMV.

As far as gcc-3.3.x is concerned, 
* rtems-4.7 and rtems-4.6 toolchains can be installed in parallel
* rtems-4.6 toolchains might not work with rtems-4.7/CVS-HEAD.
* I am using gcc-3.3.x for RTEMS-4.7/CVS only.

Finally, as most of my work throughout last year has gone into
RTEMS-4.7/CVS, I consider RTEMS-4.7/CVS to be substantially superior
than RTEMS-4.6, but of cause I am biased ;)

>  If another, which is the prefered branch 
> for on-going development.

I for one, am working on RTEMS-4.7/CVS, using binutils >= 2.14,
gcc-3.3.2, autoconf-2.59, automake-1.8.2 and more or less disregard
RTEMS-4.6.

Ralf