Re: GNAT Professional machine code listings

[Simon Wright <simon.j.wright@mac.com>]

Kilgallen@SpamCop.net (Larry Kilgallen) writes:

> The last I read in this newsgroup. ACT was inexorably opposed to the
> generation of machine code listing files.  For us this makes their
> offering a non-starter.  We need to be able to have a customer for
> our commercial product send back a stack dump from a production
> release of the software.  Then we need to be able to use that
> information to see exactly which instruction went south.  That makes
> GNAT Pro a non-starter as I see it.

A colleague in a similar situation (and more used to deep tracing than
I've had to be) had a lot of success using a disassembly (via objdump,
I think) of a GNAT powerpc-wrs-vxworks executable. The code we supply
is stripped (objcopy --strip-debug, I think) but so long as we keep
the unstripped executable all is well!

His problem was to determine exactly which component of a record
assignment was causing the Constraint_Error.

But in any case you have always been able to generate machine code
listings using the -S flag. OK, the .s files end up in the library
where the .o files would normally go, which can be confusing! (this is
with GNAT 3.16a1, don't know what later versions do).