Re: [PATCH] PowerPC, fix gdb.base/retval-large-struct.exp

[Ulrich Weigand <Ulrich.Weigand@de.ibm.com>]

Tom de Vries <tdevries@suse.de> wrote:

>My concern is older gcc compilers that do not support -fvar-tracking. 
>Those have no chance of succeeding, so we should bail out for those.
>
>OTOH, non-gcc compilers may or may not provide sufficient debug 
>information, so there's no need to bail out for those.

I'm skeptical of trying to determine in advance whether or not
the current platform supports the required entry-value records.
This not only leads to rather complicated logic in the test, but
that logic isn't actually accurate anyway ...

Note that entry-value records are only loosely associated with
the -fvar-tracking option.  As you say, compilers are free to
produce entry-value records without that option (even GCC only
requires that option at -O0).  On the other hand, there are
versions of GCC that support -fvar-tracking, but *still* do not
generate entry-value records (-fvar-tracking was added in 2004,
but entry-value records only in 2011).

I personally would be fine with just having the test fail then
(I mean, the feature actually does fail).  Or in the alternative,
recognize the scenario after the fact - the ppc back-end will
emit an error message if it cannot determine the return value
due to lack of entry-value records, so the test case might
recognize that message and downgrade from a FAIL to an
UNSUPPORTED state based on that.

Bye,
Ulrich