[libcxx] [test] Disable a test regarding error behaviour for excessively long paths on windows

[mstorsjo]

Summary

Checking for the existence of an invalidly long path name isn't
an error in itself on windows.

[mstorsjo]

Reupload, rebased closer to main to hopefully fix patch application

[• Quuxplusone]

libcxx/test/std/input.output/filesystems/fs.op.funcs/fs.op.exists/exists.pass.cpp

89

/existance/existence/

Do you have any idea why line 85 does TEST_CHECK_THROW, but none of the other test cases in this file do?

[mstorsjo]

libcxx/test/std/input.output/filesystems/fs.op.funcs/fs.op.exists/exists.pass.cpp

89

Will fix the typo.

No idea, either just random inconsistency, or whoever wrote it thought it was enough to test both error reporting paths in one place, and assume it's enough to test either of them in other cases.

[curdeius]

I don't understand why this test fails on Windows. The length of the path should have nothing to do here, the file doesn't exist even if we suppose that the filename gets truncated to 256 characters.
It sounds like some subjacent issue, no?

libcxx/test/std/input.output/filesystems/fs.op.funcs/fs.op.exists/exists.pass.cpp

89

I think we miss in lots of places the tests of throwing versions.

[mstorsjo]

In D98141#2610084, @curdeius wrote:

I don't understand why this test fails on Windows. The length of the path should have nothing to do here, the file doesn't exist even if we suppose that the filename gets truncated to 256 characters.
It sounds like some subjacent issue, no?

No, it's not about whether the file exists or not - when we have bool exists(path, error_code&), a normal nonexistent file doesn't flag any error, and returns false - but this test expects the error code to be set. So its line 94 originally or 97 in the patched version, that fails on windows. (And I don't think there's much need in keeping the rest of this test and just ifdef out that single line, as we have other tests for nonexistent files above.)

[curdeius]

OK. Thanks for making it clear. Indeed it's ok to ifdef the whole test as we have others. I had misunderstood your comment in the code. LGTM.

[mstorsjo]

Ping for second approval

[Mordante]

I see a build failure, I expect it to be a flacky failure. Can you either rerun the CI or keep an eye out after landing.
LGTM with a minor nit.

libcxx/test/std/input.output/filesystems/fs.op.funcs/fs.op.exists/exists.pass.cpp

89

s/invalidly/invalid/

[mstorsjo]

In D98141#2620073, @Mordante wrote:

I see a build failure, I expect it to be a flacky failure. Can you either rerun the CI or keep an eye out after landing.
LGTM with a minor nit.

As it's just adding an ifdef, I'd guess it's just a spurious unrelated error, but I'll keep an eye out.

libcxx/test/std/input.output/filesystems/fs.op.funcs/fs.op.exists/exists.pass.cpp

89

Thanks, will fix.

libcxx/test/std/input.output/filesystems/fs.op.funcs/fs.op.exists/exists.pass.cpp

TEST_CASE(test_exists_fails)
 
     std::error_code ec;
     TEST_CHECK(exists(file, ec) == false);
     TEST_CHECK(ec);
 
     TEST_CHECK_THROW(filesystem_error, exists(file));
 }
 
+#ifndef _WIN32
+// Checking for the existence of an invalid long path name doesn't
+// trigger errors on windows.
 TEST_CASE(test_name_too_long) {
     std::string long_name(2500, 'a');
     const path file(long_name);
 
     std::error_code ec;
     TEST_CHECK(exists(file, ec) == false);
     TEST_CHECK(ec);
 }
+#endif
 
 TEST_SUITE_END()