ptsname_r on Cygwin

ptsname_r on Cygwin

[Bruno Haible]

Hi Ken,

On Cygwin 2.9 (64-bit) I see a test failure in test-ptsname_r.c, here:

  {
    char buffer[256];
    int result;

    result = ptsname_r (-1, buffer, sizeof buffer);
    ASSERT (result != 0);  // <======================== HERE
    ASSERT (result == EBADF || result == ENOTTY);
  }

While https://www.kernel.org/doc/man-pages/online/pages/man3/ptsname_r.3.html
says that ptsname_r, upon failure, should return an error code, on Cygwin,
it returns 0 and stores the empty string in 'buffer'.

This causes 2 test failures of Gnulib tests:


FAIL: test-ptsname
==================

../../gltests/test-ptsname.c:76: assertion 'result == NULL' failed
FAIL test-ptsname.exe (exit status: 134)

FAIL: test-ptsname_r
====================

../../gltests/test-ptsname_r.c:126: assertion 'result != 0' failed
FAIL test-ptsname_r.exe (exit status: 134)


Bruno

Re: ptsname_r on Cygwin

[Bruno Haible]

> While https://www.kernel.org/doc/man-pages/online/pages/man3/ptsname_r.3.html
> says that ptsname_r, upon failure, should return an error code, on Cygwin,
> it returns 0 and stores the empty string in 'buffer'.

Let me document this issue.


2021-01-20  Bruno Haible  <bruno@clisp.org>

	doc: Document ptsname_r bug on Cygwin.
	* doc/glibc-functions/ptsname_r.texi: Mention Cygwin bug.

diff --git a/doc/glibc-functions/ptsname_r.texi b/doc/glibc-functions/ptsname_r.texi
index 0bfd05b..a79e751 100644
--- a/doc/glibc-functions/ptsname_r.texi
+++ b/doc/glibc-functions/ptsname_r.texi
@@ -25,13 +25,17 @@ Mac OS X 10.5, FreeBSD 6.0, NetBSD 5.0, OpenBSD 6.7, Minix 3.1.8,
 AIX 5.1, HP-UX 11, IRIX 6.5, Solaris 11.4, Cygwin 1.7.9, mingw,
 MSVC 14.
 @item
-When this functions fails, it returns -1 instead of the error code
+When this function fails, it returns -1 instead of the error code
 on some platforms:
 Mac OS X 10.13, Android 4.3.
 @end itemize
 
 Portability problems not fixed by Gnulib:
 @itemize
+@item
+When this function fails, it returns 0 and stores an empty string as result
+on some platforms:
+Cygwin 2.9.
 @end itemize
 
 Note: Portable programs should expect to find the error code as the

Re: ptsname_r on Cygwin

[Ken Brown]

Hi Bruno,

On 1/20/2021 1:31 AM, Bruno Haible wrote:
> Hi Ken,
> 
> On Cygwin 2.9 (64-bit) I see a test failure in test-ptsname_r.c, here:
> 
>    {
>      char buffer[256];
>      int result;
> 
>      result = ptsname_r (-1, buffer, sizeof buffer);
>      ASSERT (result != 0);  // <======================== HERE
>      ASSERT (result == EBADF || result == ENOTTY);
>    }
> 
> While https://www.kernel.org/doc/man-pages/online/pages/man3/ptsname_r.3.html
> says that ptsname_r, upon failure, should return an error code, on Cygwin,
> it returns 0 and stores the empty string in 'buffer'.
> 
> This causes 2 test failures of Gnulib tests:
> 
> 
> FAIL: test-ptsname
> ==================
> 
> ../../gltests/test-ptsname.c:76: assertion 'result == NULL' failed
> FAIL test-ptsname.exe (exit status: 134)
> 
> FAIL: test-ptsname_r
> ====================
> 
> ../../gltests/test-ptsname_r.c:126: assertion 'result != 0' failed
> FAIL test-ptsname_r.exe (exit status: 134)

Thanks for the report.  I've submitted a patch that fixes this:

   https://cygwin.com/pipermail/cygwin-patches/2021q1/010995.html

Ken

Re: ptsname_r on Cygwin

[Bruno Haible]

Ken Brown wrote:
> Thanks for the report.  I've submitted a patch that fixes this:
> 
>    https://cygwin.com/pipermail/cygwin-patches/2021q1/010995.html

Thanks. EBADF is the more "proper" errno value here, for fd < 0. But
glibc produces errno ENOTTY in this case; so it should be acceptable too.

Bruno

Re: ptsname_r on Cygwin

[Ken Brown]

On 1/21/2021 3:26 PM, Bruno Haible wrote:
> Ken Brown wrote:
>> Thanks for the report.  I've submitted a patch that fixes this:
>>
>>     https://cygwin.com/pipermail/cygwin-patches/2021q1/010995.html
> 
> Thanks. EBADF is the more "proper" errno value here, for fd < 0. But
> glibc produces errno ENOTTY in this case; so it should be acceptable too.

I thought EBADF made more sense too, but Cygwin generally tries to emulate 
Linux.  We'll see what Corinna thinks.

Ken

Re: ptsname_r on Cygwin

[Ken Brown]

On 1/21/2021 5:45 PM, Ken Brown wrote:
> On 1/21/2021 3:26 PM, Bruno Haible wrote:
>> Ken Brown wrote:
>>> Thanks for the report.  I've submitted a patch that fixes this:
>>>
>>>     https://cygwin.com/pipermail/cygwin-patches/2021q1/010995.html
>>
>> Thanks. EBADF is the more "proper" errno value here, for fd < 0. But
>> glibc produces errno ENOTTY in this case; so it should be acceptable too.
> 
> I thought EBADF made more sense too, but Cygwin generally tries to emulate 
> Linux.  We'll see what Corinna thinks.

Corinna agreed with you but left it up to me, so I went with EBADF.

Thanks.

Ken