The following series of patches fixes a number of test failures when running
the GCC DejaGNU testsuite for msp430-elf.
The raw output from contrib/compare_tests is a bit misleading for some tests,
as lines have been added to the source code, changing line numbers for
dg-warning/dg-error tests. I verified there are no regressions
or new failures for x86_64-pc-linux-gnu (gcc, g++), avr (gcc) and msp430-elf
with the small and large memory model (gcc, g++).
For the msp430 small memory model (default) here are the summary results:
BEFORE:
=== gcc Summary ===
# of expected passes 90663
# of unexpected failures 447
# of unexpected successes 6
# of expected failures 287
# of unresolved testcases 93
# of unsupported tests 4171
=== g++ Summary ===
# of expected passes 99772
# of unexpected failures 2673
# of expected failures 433
# of unresolved testcases 1812
# of unsupported tests 5862
AFTER:
=== gcc Summary ===
# of expected passes 90750
# of unexpected failures 150
# of unexpected successes 3
# of expected failures 295
# of unresolved testcases 14
# of unsupported tests 4313
=== g++ Summary ===
# of expected passes 99804
# of unexpected failures 706
# of expected failures 436
# of unresolved testcases 46
# of unsupported tests 7778
For the -mlarge configuration here are the summary results:
BEFORE:
=== gcc Summary ===
# of expected passes 90451
# of unexpected failures 590
# of unexpected successes 7
# of expected failures 286
# of unresolved testcases 162
# of unsupported tests 4172
=== g++ Summary ===
# of expected passes 99509
# of unexpected failures 2885
# of expected failures 433
# of unresolved testcases 1875
# of unsupported tests 5862
AFTER:
=== gcc Summary ===
# of expected passes 90641
# of unexpected failures 154
# of unexpected successes 3
# of expected failures 293
# of unresolved testcases 32
# of unsupported tests 4355
=== g++ Summary ===
# of expected passes 99529
# of unexpected failures 949
# of expected failures 436
# of unresolved testcases 128
# of unsupported tests 7759
The "before" results do not include the "obvious" testsuites changes committed
in r265924, r265926, r265927, r265928,
(https://gcc.gnu.org/ml/gcc-patches/2018-11/msg00618.html)
but the "after" results do.
Patch 1 tweaks dg directives in tests specifically for msp430. Many of
these are extensions to existing target selectors in dg directives.
Patch 2 fixes issues finding libssp when linking tests or checking for
fstack_protector support.
Patch 3 sets up require-effective-target directives for tests which
require the compilation of large arrays.
Targets which have 16-bit or 20-bit size_t fail to compile tests with large
arrays designed to test 32-bit or 64-bit behaviour. Rather than enumerating
another target to skip, I've replaced the target selector in some tests with
a size checking procedure:
- size20plus (new)
- size32plus
size20plus checks to see if a 16-bit structure/array size is supported,
similarly to how the existing size32plus checks to see if a 24-bit
structure/array size is supported,
Patch 4 fixes tests when int is 16-bits by default.
Patch 5 deals with ISO C errors emitted by tests when the large memory model is
used. size_t and ptrdiff_t are __int20 with -mlarge, and if the test is
compiled with -pedantic-errors and -std=* or -ansi, then use of these types
causes an error of the form:
ISO C does not support __int20 types
I fixed this by adding dg-prune-output directives to tests which cause this
error.
Alternatively, I considered adding typedefs preceded by __extension__ to fix
these errors, but in many cases __SIZE_TYPE__ is directly used so replacing all
these with a new typedef'd type changes the code in more places, in some cases
changing the offset for dg-warning or dg-error directives. Changing the line
numbers for dg-warning/dg-error adds further manual steps to comparing
testresults and as these are generic tests I wanted to minimize the effect on
the testresults for other targets.
Patch 6 fixes tests expecting printf float support for targets which have been
configured with "newlib-nano-formatted-io". When newlib is configured in this
way, float printf is enabled at build time by registering _printf_float as an
undefined symbol.
Patch 7 fixes tests for msp430-elf in the large memory model.
Ok for trunk?
