Closed kangaroo closed 4 years ago
Some more information about the store for my memory:
Fixed by https://gist.github.com/kangaroo/bd1f95684518b0f70ad5
I need to double check the ABI docs in the morning to ensure this is the correct behavior.
Maybe long double
or long long
Following links maybe help:
cc @sergiy-k I havn't dug more, but this might be a sub-cause of https://github.com/dotnet/coreclr/issues/161 so I'm going to tag you and shelve this until you submit a PR for https://github.com/dotnet/coreclr/issues/161
@kangaroo I have assigned this issue to myself. I thought you have already closed https://github.com/dotnet/coreclr/issues/161 since you have tested my change and the corresponding PR (https://github.com/dotnet/coreclr/pull/258) has been already merged.
By the way, on Mac (release build only) 14 fwprintf tests fail. But I have not checked whether these failures are related to the problem in PAL_printf_arg_remover.
c_runtime/fwprintf/test10/paltest_fwprintf_test10. Exit code: 1 c_runtime/fwprintf/test11/paltest_fwprintf_test11. Exit code: 1 c_runtime/fwprintf/test12/paltest_fwprintf_test12. Exit code: 1 c_runtime/fwprintf/test13/paltest_fwprintf_test13. Exit code: 1 c_runtime/fwprintf/test14/paltest_fwprintf_test14. Exit code: 1 c_runtime/fwprintf/test15/paltest_fwprintf_test15. Exit code: 1 c_runtime/fwprintf/test16/paltest_fwprintf_test16. Exit code: 1 c_runtime/fwprintf/test17/paltest_fwprintf_test17. Exit code: 1 c_runtime/fwprintf/test18/paltest_fwprintf_test18. Exit code: 1 c_runtime/fwprintf/test3/paltest_fwprintf_test3. Exit code: 1 c_runtime/fwprintf/test4/paltest_fwprintf_test4. Exit code: 1 c_runtime/fwprintf/test6/paltest_fwprintf_test6. Exit code: 1 c_runtime/fwprintf/test8/paltest_fwprintf_test8. Exit code: 1 c_runtime/fwprintf/test9/paltest_fwprintf_test9. Exit code: 1
@sergiy-k dotnet/coreclr#161 is closed but I haven't re-tested this yet. It wasn't exposed by a PAL test, but by the k-nucleotide managed executable. I'll see if I have some time this weekend to confirm this one, and look at the fwprintf ones.
Its more mis-use of va_lists all around. I'm cleaning it up.
@sergiy-k Fixed in https://github.com/dotnet/coreclr/pull/271
This is fixed. Can someone close it out please?
Its friday night and I'm tired, so I'm going to file this to document it.
k-nucleotide.cs from the language shootout is currently causing a fault on OSX. The stack trace is as follows:
The reason is in CoreVsnprintf the call to PAL_printf_arg_remover is clobbering the upper 32-bits of ap. Commenting out line 151 works around the issue, but obviously isn't a proper fix.
PAL_printf_arg_remover is:
More direct context: