[ECOS] Question on GDB stubs startup type.

[ECOS] Question on GDB stubs startup type.

[Ling Su]

Hi, Dear All,

I posted following message yesterday, but got no answer, could anyone take a
look at it. Sorry for the bothering.

I am porting GDB stubs onto little endian VRC4375 based on VRC4373 codes.
With the help of PMON monitor, now I can make a ROM startup GDBstubs monitor
working. It can download code can single step debug. Of course, I have have
trouble to run all the test senarios. I am debugging it now. Since I port
this stubs based on eCos-1.3.1, I am looking at the memory layout between it
and the CVS repository, I found there are some difference for vrc4373. The
startup for GDBstubs has been changed to ROMRAM instead of ROM, I wonder
what's the difference, can anyone kindly offer my some explaination?

The most tricky thing is, when I observe the binary loader provided in
eCos-1.3.1 release, I found it is ROMRAM startup actually, which is
obviously different from the source code release. So I guess RedHat use a
different set of code to produce the binary stubs loader, is that true?

Anyway, I thought ROMRAM startup should be the correct one, so I go ahead to
produce a ROMRAM startup version, to my surprise, this time the monitor is
running, I can see its running using a Logical Analyzer, but the behavior is
some different from ROM startup one, at least it won't response my GDB
inqury any more. :-( To be more procise, the write cyclye from PCI bus to
peripherials is much longer, it always try 2 or 3 times for a successful
write. It really puzzled me. I hope someone ever writing/building the
vrc4373 stubs can give me some hint on those problems! Espeically those guys
from RedHat(of course, I knew all of you are so kind and willing to offer
assisatances.:)), since I am a little bit afraid of the inconsistence
between the source and the binary in eCos-1.3.1 release. Thanks very much!

Best Regards,
-Ling