XCOFF ld testsuite fixes

[Jan Beulich]

On 13.07.2020 12:15, Alan Modra wrote:
> On Mon, Jul 13, 2020 at 09:14:06AM +0200, Jan Beulich wrote:
>> On 07.07.2020 10:56, Alan Modra via Binutils wrote:
>>> 	* testsuite/ld-scripts/align.exp: Don't exclude xcoff.  Pass
>>> 	-bnogc ld option for xcoff.
>>> 	* testsuite/ld-scripts/provide.exp: Likewise.
>>> 	* testsuite/ld-scripts/data.exp: Pass -bnogc ld option for xcoff.
>>> 	* testsuite/ld-scripts/default-script.exp: Likewise.
>>> 	* testsuite/ld-scripts/defined.exp: Likewise.
>>> 	* testsuite/ld-scripts/empty-address.exp: Likewise.
>>> 	* testsuite/ld-scripts/expr.exp: Likewise.
>>> 	* testsuite/ld-scripts/include.exp: Likewise.
>>> 	* testsuite/ld-scripts/script.exp: Likewise.
>>> 	* testsuite/ld-scripts/assign-loc.d: Don't exclude xcoff.
>>> 	* testsuite/ld-scripts/defined3.d: Likewise.
>>> 	* testsuite/ld-scripts/defined4.d: Likewise.
>>> 	* testsuite/ld-scripts/pr18963.d: Likewise.
>>> 	* testsuite/ld-scripts/sane1.d: Likewise.
>>> 	* testsuite/ld-scripts/segment-start.d: Likewise.
>>> 	* testsuite/ld-scripts/include-1.d: Likewise, and relax text vma.
>>> 	* testsuite/ld-scripts/defined5.d: Update xfail and comment.
>>> 	* testsuite/ld-scripts/defined5.s: Tweak "defined" to be at
>>> 	non-zero section offset.
>>> 	* testsuite/ld-scripts/fill16.d: xfail for xcoff.
>>> 	* testsuite/ld-scripts/provide-2.d: Accept more symbols.
>>> 	* testsuite/ld-scripts/provide-4.d: Likewise.
>>> 	* testsuite/ld-scripts/provide-5.d: Likewise.
>>> 	* testsuite/ld-scripts/provide-6.d: Likewise.
>>> 	* testsuite/ld-scripts/provide-7.d: Likewise.
>>> 	* testsuite/ld-scripts/align.t: Accept xcoff mapped .text and .data.
>>> 	* testsuite/ld-scripts/defined3.t: Likewise.
>>> 	* testsuite/ld-scripts/defined4.t: Likewise.
>>> 	* testsuite/ld-scripts/defined5.t: Likewise.
>>> 	* testsuite/ld-scripts/fill.t: Likewise.
>>> 	* testsuite/ld-scripts/include-subdata.t: Likewise.
>>> 	* testsuite/ld-scripts/provide-1.t: Likewise.
>>> 	* testsuite/ld-scripts/provide-2.t: Likewise.
>>> 	* testsuite/ld-scripts/provide-3.t: Likewise.
>>> 	* testsuite/ld-scripts/provide-4.t: Likewise.
>>> 	* testsuite/ld-scripts/provide-5.t: Likewise.
>>> 	* testsuite/ld-scripts/provide-6.t: Likewise.
>>> 	* testsuite/ld-scripts/provide-7.t: Likewise.
>>> 	* testsuite/ld-scripts/provide-8.t: Likewise.
>>
>> Alan,
>>
>> presumably as of this change I'm observing, for x86_64-cygwin and
>> besides many FAIL-s no longer occurring there,
>>
>> XPASS: ld-scripts/provide-6
>> XPASS: ld-scripts/provide-8
>>
>> Was it perhaps meant for the respective xfail attributes of the
>> tests to be dropped? Or am I running in so much different an env
>> (cross build on 64-bit Linux) than you?
> 
> I regularly build a cross to x86_64-w64-mingw32 but not to
> x86_64-cygwin.  So I didn't see this difference in test results.
> 
> Of course, the test just having "xfail: x86_64-*-cygwin" isn't good,
> since presumably if a test fails on x86_64-cygwin then it also fails
> on x86_64-*-pe* and x86_64-*-mingw*.
> 
> Does the following work for you?

Yes, it makes the XPASS-es go away as expected. It looks to have
other effects too, though - not sure if that's expected. For
cygwin from

# of expected passes            118
# of unexpected successes       2
# of expected failures          2
# of untested testcases         21
# of unsupported tests          53

to

# of expected passes            121
# of expected failures          6
# of untested testcases         21
# of unsupported tests          52

and for mingw64 from

# of expected passes            124
# of expected failures          2
# of untested testcases         20

to

# of expected passes            124
# of expected failures          6
# of untested testcases         20

In any event - thanks for the quick fix,
Jan