On 8/8/24 10:28 PM, Joern Wolfgang Rennecke wrote:
I'm not quite that sure about the general applicability of these, as
these depend somewhat on code size. Although there might be something
we can prove about a minimum frame size for one test or the other at -O0.
I also tried to add
/* { dg-skip-if "memory tight" { !size20plus } { "-O3" } } */
and
/* { dg-require-effective-target size20plus } */
to gcc.c-torture/execute/builtins/snprintf-chk.c , but to no avail.
2024-07-15 Joern Rennecke<joern.renne...@riscy-ip.com>
testsuite/
* gcc.c-torture/execute/20021120-1.c: Skip if not size20plus of -Os.
* gcc.dg/fixed-point/convert-float-4.c: Require size20plus.
* gcc.dg/torture/pr112282.c: Skip if -O0 unless size20plus.
* g++.dg/lookup/pr21802.C: Require size20plus.
I don't have any of these marked as problematical in my tester, but that
mostly just means they haven't been seen as sometimes working and
sometimes not on the embedded targets. It's also the case that I've
twiddled a few targets to have more address space to facilitate
consistent test results when we are short on address space.
If you're seeing them fail on a target with limited address space, then
by all means to ahead add the appropriate dg-skip-if lines. The vast
majority of times the testsuite is run these tests will still be run.
For the affected target(s) I would say the consistency in results is
probably more important than running these 4 tests.
So, yea, go ahead and commit to the trunk. If you run into others of a
similar nature, go ahead and commit those fixes as well.
jeff
