[Bug 1000184] Ecos kernel data corrupted

[bugzilla at ecoscentric dot com]

http://bugs.ecos.sourceware.org/show_bug.cgi?id=1000184





------- Additional Comments From erik@dd.nec.com.au  2005-01-07 07:43 -------
Hi Jonathan,

My fully working test case, essentially the same as your foo.c, runs on the same
board, with the same BSP and libtarget.a, as Kori uses. It has 4 threads happily
playing ping-pong via mboxes, with CYG_KERNEL_SYNC_MBOX_QUEUE_SIZE==50. It
differs from foo.c primarily in using LEDs for output, because we're not using
diag_printf. I've also breakpointed through 50 writes & 50 reads, on a single
thread, watching the queue build and shrink in RAM, without corruption, using an
ICE.

I'm not too happy about taking up your time at this stage, 'coz I'm not yet
convinced we have a handle on a bug in eCos. There may be some inconspicuous
initialisation we've missed, like the cyg_hal_invoke_constructors() I found down
in the "adder" board's vectors.S. Whatever the problem, it is clearly not
directly related to queue size. We'll report what we find.

Many Thanks,
Erik



------- You are receiving this mail because: -------
You are the QA contact for the bug, or are watching the QA contact.