On Fri, 2020-05-15 at 18:26 +0100, Jozef Lawrynowicz wrote:
> The attached patch fixes many GCC and G++ tests for 16-bit targets. These
> targets can have the following properties:
> - "int", "size_t", "ptrdiff_t", "void *" are 16-bit types
> - sizeof(int) == sizeof(short)
>
> These properties cause problems for a number of tests in the testsuite,
> where int is often assumed to be a 32-bit type (and the test relies on that
> property to run as expected or to not generate any unexpected warnings).
> Other failures occur when arrays larger than what is supported for 16-bit
> targets are declared, or the "short" type is expected to be smaller than
> "int".
>
> The fixes fall into a few different categories:
> - Explicitly defining a 32-bit int type and using that in place of "int" or
> "unsigned int".
> - Skipping tests which rely on arrays sized larger than what is supported by
> the
> target.
> - Adding or adjusting dg-{warning,error} directives
> - Other testcase specific adjustments.
>
> I've successfully regtested the patch on x86_64-pc-linux-gnu and msp430-elf in
> the default, -mlarge and -mcpu=msp430 configurations.
>
> There are no absolute changes to testresults on x86_64-pc-linux-gnu, but
> because
> of changes to the line number of dg-{warning,message,error,etc} directives,
> the
> results comparison shows:
> UNTESTED->FAIL: 8 tests
> UNTESTED->PASS: 986 tests
> PASS->UNTESTED: 986 tests
> FAIL->UNTESTED: 8 tests
>
> Across the 3 configurations for msp430, the total is:
> PASS->FAIL: 0 tests
> FAIL->PASS: 1108 tests
> UNTESTED->FAIL: 24 tests
> UNTESTED->PASS: 2924 tests
> FAIL->UNTESTED: 807 tests
> PASS->UNTESTED: 2564 tests
>
> Ok for trunk?
>
> What about for gcc-10 branch? It would be nice to clean up the testresults for
> MSP430, but I understand if it is undesirable to cause these line number
> changes
> to appear on the stable branch.
OK for the trunk. I suspect the release managers wouldn't want it for the
branch
though.
jeff
