Release plan for RTEMS 4.11?

Ralf Corsepius ralf.corsepius at rtems.org
Tue Apr 3 08:42:33 UTC 2012


On 04/03/2012 09:58 AM, Sebastian Huber wrote:
> Hi,
>
> which open points do we have for RTEMS 4.11? Do we have a time line?

No. Common sense would be to have

1. A shorter (severals weeks) "no new features", "bug-fix-only", code 
consolidation phase.

2. If being satisfied with the results, a fork and subsequently followed 
by a deep-freeze, "hard-testing"-only phase would follow.

Experience tells, there are different visions of "being satisfied" with 
the code.

[You can already smell it: You seem to be considering RTEMS 4.11 to be 
in "close to release shape" - I could not disagree more :( ]


Experience also tells, the RTEMS core devs will not care about the 
"production" release code once having forked, because they will consider 
this to be "boring", "tedious", "non-interesting" and "duplication of work".


>  From my side there are some file system issues outstanding.
>
> 1. The lseek() mechanic should change. We should get rid of the size
> field in the rtems_libio_t structure.
>
> 2. We should remove the unused sem field in the rtems_libio_t structure.
> This reduces per file descriptor overhead by roughly 128 Bytes.
>
> 3. lseek() and ftruncate() should be made POSIX conform.

 From my POV, RTEMS-4.11 is far from being in release shape, because too 
many changes have found their way into RTEMS, which have significantly 
destabilized RTEMS and have broken things.

Just one example: With today's git, ca. 50% of all powerpc bsps do not 
even build  (for various reasons).

> Regarding GCC 4.7 I am not sure if this is the right candidate for the
> RTEMS 4.11 release. We tested mostly only on GCC 4.6.

I am strongly in favor of switching to GCC-4.7.

However, the sad truth is, GCC-4.7.0 doesn't build for ca. 30% of all 
RTEMS targets and is suffering from other open issues on ca. 20% more 
[1]. It builds and seems to work flawlessly for the remaining 50% of 
targets.

Ralf

[1] avr, m32c and sparc64 ICE.
sparc and sh are having major integration probs, arm and powerpc are 
having smaller integration probs.






More information about the devel mailing list