[Openmcl-devel] ARM testing

[Gary Byers]

On Thu, 20 Jan 2011, David Brown wrote:

> On Wed, Jan 19 2011, Gary Byers wrote:
>
>> The (small) trick:
>>
>>  - checkout the trunk
>>  - rebuild the lisp kernel
>>  - rebuild the image, using the new kernel
>>  - rebuild the image again
>
> I did the rebuild on the trunk.  The build of the image and boot took a
> _very_ long time.  When trying to load the boot image, I'm getting a bus
> error with what looks like a stack overflow.  I got it to continue with
> a sufficiently large -S argument to loading the boot image.
>
> The second build was much faster.  I did still have to manually do the
> bootstrap to prevent the stack overflow.
>

I don't think that I saw any of that, but I cross-compiled when building
for the Nexus and my other ARM machines weren't affected by the conditional
trap issue.

If the problem persists for you, that'd suggest that there's something
else going on.  (There's still some flakiness in the ARM port in general -
automated builds fail with weird errors some noticable percentage of
the time - but those failures generally aren't reproducible.)

I don't have a good guess as to what would be causing excessive (but bounded)
stack usage.

> Looking good, though.
>
> The rebuilt image is about 4MB smaller than the old one, should this
> surprise me?

That has to do with how the source location recording stuff
works.   The short version is that options in my init file
cause it to save the full source text of every definition.

It's kind of amazing that that's only 4MB, but that accounts
for the size difference.


>
> David
>
>