Re: Why require SLOW_BUT_NO_HACKS for stubs?

2012-06-24 Thread Paul Eggert
On 06/24/2012 03:42 PM, John Spencer wrote: > anything is better than a failed build. Isn't this discussion moot now, with respect to musl? That is, I thought the problem with musl and gnulib is fixed, so we don't have a failed build now. If this discussion is about what to do with some other new

Re: ptsname_r on OSF/1

2012-06-24 Thread Bruno Haible
> 2012-06-24 Bruno Haible > > ptsname_r: Make it consistent with ptsname on OSF/1. > * lib/ptsname_r.c (__ptsname_r): Add a different implementation for > OSF/1. Nearly the same code works also on AIX. Unfortunately, it does not fix the problem that close() hangs when applied

Re: Why require SLOW_BUT_NO_HACKS for stubs?

2012-06-24 Thread John Spencer
On Sun, 24 Jun 2012 04:05:39 -0700, Bruno Haible wrote: Paolo Bonzini wrote: > >> > The test as it stands is "error out on unsupported platforms unless > >> > user specifies to use slow method". > >> > My proposal is "On unsupported platforms, use the slow method instead >

ptsname_r on OSF/1

2012-06-24 Thread Bruno Haible
On OSF/1, gnulib's ptsname_r function returns a file name like "/dev/ttyp5". But ptsname() returns something like "/dev/pts/5". How does it do this? 'truss' reveals it: ioctl(3, 0x20007447, 0x)= 7340037 = 0x75 stat("/dev/pts/5", 0x00011FFFBA70) = 0 To

ttyname_r on OSF/1

2012-06-24 Thread Bruno Haible
On OSF/1, I'm seeing this test failure in an *interactive* "make check" of a ptsname_r testdir: test-ptsname_r.c:193: assertion failed FAIL: test-ptsname_r It turns out the cause is a bug in gnulib's ttyname_r: It fails to add a trailing NUL byte to the output string. This fixes it. 2012-06

ptsname_r on Solaris

2012-06-24 Thread Bruno Haible
When running the test for module 'ptsname_r' on Solaris *interactively*, I'm seeing this error: test-ptsname_r.c:163: assertion failed FAIL: test-ptsname_r Solaris needs a different approach for ptsname_r. The (not reentrant) ptsname function from the OS does the following system calls, obser

Re: ptsname_r test

2012-06-24 Thread Bruno Haible
Similarly, on mingw I'm seeing this test failure: test-ptsname_r.c:117: assertion failed FAIL: test-ptsname_r.exe and on MSVC 9, it just crashes: FAIL: test-ptsname_r.exe Again, the reason lies in the isatty() function. This fixes it. 2012-06-24 Bruno Haible ptsname_r: Fix t

ptsname_r test

2012-06-24 Thread Bruno Haible
On Solaris 10 and IRIX 6.5, I'm seeing this test failure: test-ptsname_r.c:117: assertion failed FAIL: test-ptsname_r The reason is that isatty() has not set errno. This fixes it. 2012-06-24 Bruno Haible ptsname_r: Fix test failures on IRIX, Solaris. * m4/ptsname_r.m4 (gl_PR

ptsname test

2012-06-24 Thread Bruno Haible
On all platforms, ptsname() sets errno when the argument is an invalid file descriptor. Even through POSIX does not require it [1], it's safe to check it. [1] http://pubs.opengroup.org/onlinepubs/9699919799/functions/ptsname.html 2012-06-24 Bruno Haible ptsname test: Extend test.

Re: Why require SLOW_BUT_NO_HACKS for stubs?

2012-06-24 Thread Bruno Haible
Paolo Bonzini wrote: > >> > The test as it stands is "error out on unsupported platforms unless > >> > user specifies to use slow method". > >> > My proposal is "On unsupported platforms, use the slow method instead > >> > of erroring out." > > > > If we did this, nobody would report to bug-gnulib