Re: [PATCH 0/2] fix undefined reference to `__ssputws_r'

Hi again,

I do not have time to dig into this right now, but these are the test 
cases that start to fail after applying the 2 patches suggested by Hau Hsu:

FAIL: 27_io/objects/wchar_t/12048-1.cc  -std=gnu++17 execution test
FAIL: 27_io/objects/wchar_t/12048-2.cc  -std=gnu++17 execution test
FAIL: 27_io/objects/wchar_t/12048-3.cc  -std=gnu++17 execution test
FAIL: 27_io/objects/wchar_t/12048-4.cc  -std=gnu++17 execution test
FAIL: 27_io/objects/wchar_t/2.cc  -std=gnu++17 output pattern test
FAIL: 27_io/objects/wchar_t/9662.cc  -std=gnu++17 execution test
FAIL: ext/stdio_sync_filebuf/wchar_t/1.cc  -std=gnu++17 execution test
FAIL: ext/stdio_sync_filebuf/wchar_t/12948-1.cc  -std=gnu++17 execution test
FAIL: ext/stdio_sync_filebuf/wchar_t/12948-2.cc  -std=gnu++17 execution test
FAIL: ext/stdio_sync_filebuf/wchar_t/12948-3.cc  -std=gnu++17 execution test
FAIL: ext/stdio_sync_filebuf/wchar_t/12948-4.cc  -std=gnu++17 execution test

This was checked with basepoints/gcc-15-1852-g320c2ed4d2b and newlib 
1339af44679aee0895fe311cfad89d38cfc2b919 built for arm-none-eabi.

Kind regards,
Torbjörn


On 2024-07-12 10:07, Torbjorn SVENSSON wrote:
> Hi,
> 
> I've looked at the patch and I think it's fine.
> I am currently building a toolchain with the patch series applied and 
> will run the tests on it. I'll get back with the result early next week.
> 
> The reason why I didn't go this route before was that I was afraid that 
> there would be more functions that would be missing. Hopefully, the 
> tests will show that this is not the case.
> 
> Kind regards,
> Torbjörn
> 
> On 2024-07-09 12:48, Corinna Vinschen wrote:
>> @Torbjörn,
>>
>> would you mind to take a look?  Any input on this?
>>
>> FWIW< this looks right to me.
>>
>>
>> Thanks,
>> Corinna
>>
>>
>> On Jul  8 12:01, Hau Hsu wrote:
>>> *Issue*
>>> When compile a simple C++ hello world, we got undefined reference to
>>> `__ssputws_r' error with NEWLIB_NANO_FORMATTED_IO after upgrading to
>>> newlib-4.4 in risc-v toolchain:
>>>
>>> svfwprintf.c:(.text._svfwprintf_r+0x8e): undefined reference to
>>> `__ssputws_r'
>>>
>>> *Cause*
>>> This is because commit
>>> "v{fs}printf/v{fs}wprintf: create external output helpers"
>>> <https://cygwin.com/git/?p=newlib- 
>>> cygwin.git;a=commit;h=61ccd3f94f92bcfc0940f0595ea5b3b72bce3c6d>
>>> introduces a new file ssputws_r.c for a wide-char-oriented helper 
>>> function.
>>> But this file will not be built when newlib_nano_formated_io is enabled.
>>>
>>> The problem is that the function __ssputws_r() is used by svfwprintf(),
>>> which is compiled no matter whether newlib_nano_formated_io is enabled.
>>>
>>> *Fix*
>>> This patch fix the issue by building ssputws_r.c regardless of whether
>>> newlib_nano_formated_io is enabled.
>>>
>>> *Problem of the previous fix*
>>> Note that we also reverts a previouls patch that tries to fix the same
>>> link error:
>>> <https://cygwin.com/git/?p=newlib- 
>>> cygwin.git;a=commitdiff;h=3b97a5ec67a5a52c130158bb143949cd842de305>
>>> Because we got other errors after applying this patch:
>>>
>>> undefined reference to `putwc'
>>> ...
>>> undefined reference to `getwc'
>>> ...
>>> undefined reference to `swprintf'
>>>
>>> The previous patch removes wchar_t functions from
>>> NEWLIB_NANO_FORMATTED_IO, but in the newlib/README says:
>>>
>>>> --enable-newlib-nano-formatted-io
>>>>    This option does not affect wide-char formatted I/O functions
>>>
>>> Thus I think we shouldn't remove wchar_t functions from
>>> newlib_nano_formated_io.
>>>
>>> Similar bug report:
>>> * https://inbox.sourceware.org/newlib/ 
>>> [email protected]/t/
>>>
>>>
>>>
>>> Hau Hsu (2):
>>>    Revert "libc/stdio: Remove wchar_t functions from
>>>      NEWLIB_NANO_FORMATTED_IO"
>>>    libc/stdio: fix undefined reference to `__ssputws_r'
>>>
>>>   newlib/Makefile.in             | 1004 ++++++++++++++++----------------
>>>   newlib/libc/stdio/Makefile.inc |   54 +-
>>>   2 files changed, 528 insertions(+), 530 deletions(-)
>>>
>>> -- 
>>> 2.31.1
>>