Re: netdb.h, getaddrinfo

Bruno Haible writes: > Hi Simon, > > A week ago I proposed: >> > I propose to generate netdb.h always, like we do >> > with stdlib.h and unistd.h for 3 years and for many others for >> > 2 months. Then we can also add GNULIB_POSIXCHECK advice to . > > Here's the second part: The patch to add GNUL

Re: testdir without unportable tests?

Simon Josefsson writes: > Bruno Haible writes: > >> I'm applying these two patches. > > Thanks, it should likely solve the problem too -- I'm running a Windows > cygwin build to test it now. Yep: http://autobuild.josefsson.org/gnulib/log-20100425225612508.txt Disabling the longrunning che

Re: [GNU Bison 2.4.2] testsuite: Failing tests of bison 2.4.2 on Solaris 8+9 w/Sun Studio 11+12

On Sun, 25 Apr 2010, Dagobert Michelsen wrote: > However, I had the impression you wanted to > implement different checking for cc and CC? I thought that would be the most general solution, but the consensus appears to be that gnulib should not have to handle this. On Mon, 26 Apr 2010, Bruno Ha

Re: netdb.h, getaddrinfo

Hi Simon, A week ago I proposed: > > I propose to generate netdb.h always, like we do > > with stdlib.h and unistd.h for 3 years and for many others for > > 2 months. Then we can also add GNULIB_POSIXCHECK advice to . Here's the second part: The patch to add GNULIB_POSIXCHECK advice. OK to commi

Re: [GNU Bison 2.4.2] testsuite: Failing tests of bison 2.4.2 on Solaris 8+9 w/Sun Studio 11+12

To summarize so far: - The problem arises in /opt/csw/include/getopt.h, which is non-standard on Solaris. Not a problem gnulib is to be blamed for. - It arises in C++ mode only. In C mode the header file works fine. - Joel Denny writes: > it starts to feel like /opt/csw/include/getopt.h's ap

Re: ttyname_r on Solaris

> 2010-04-25 Bruno Haible > > ttyname_r: Make it work on MacOS X 10.4 and Solaris 10. On Solaris, there are two additional complications: 1) When combined with the 'extensions' module, _POSIX_PTHREAD_SEMANTICS gets defined, and the system function that returns 'int' becomes visibl

Re: ttyname_r on Solaris

> Should we define _POSIX_PTHREAD_SEMANTICS? > > Should we document in the INSTALL file that on Solaris, > _POSIX_PTHREAD_SEMANTICS > should be defined? (Like we do for _XOPEN_SOURCE on HP-UX.) Actually, the gnulib 'extensions' module already ensures that _POSIX_PTHREAD_SEMANTICS gets defined (a

ptsname: fix bug inherited from glibc

This fixes a bug relating to the invocation of ttyname_r. I've reported it as glibc bug . 2010-04-25 Bruno Haible ptsname: Fix misuse of ttyname_r. * lib/ptsname.c (__ptsname_r): Use __ttyname_r's return value instead

Re: ttyname_r on Mac

Hi Simon, > There appears to be a ttyname_r signature bug, see: > http://autobuild.josefsson.org/gnulib/log-201004250717723515000.txt > > g++ -DHAVE_CONFIG_H -I. -DGNULIB_STRICT_CHECKING=1 -I. -I. -I.. -I./.. > -I../gllib -I./../gllib -I/opt/local/include -MT test-fcntl-h-c++.o -MD -MP > -MF

Re: [GNU Bison 2.4.2] testsuite: Failing tests of bison 2.4.2 on Solaris 8+9 w/Sun Studio 11+12

Hi Joel, Am 23.04.2010 um 05:24 schrieb Joel E. Denny: > On Wed, 21 Apr 2010, Paolo Bonzini wrote: >> Having OpenCSW regenerate the libgnugetopt package with "gnulib-tool >> --create-testdir --dir=libgnugetopt getopt-gnu" and dropping support >> for the old package seems easiest, and does not requ

ttyname_r on Solaris

Simon Josefsson wrote: > There appears to be a ttyname_r signature bug, see: > http://autobuild.josefsson.org/gnulib/log-201004250717723515000.txt > > g++ -DHAVE_CONFIG_H -I. -DGNULIB_STRICT_CHECKING=1 -I. -I. -I.. -I./.. > -I../gllib -I./../gllib -I/opt/local/include -MT test-fcntl-h-c++.o -M

Re: testdir without unportable tests?

Bruno Haible writes: > I'm applying these two patches. Thanks, it should likely solve the problem too -- I'm running a Windows cygwin build to test it now. /Simon

Re: testdir without unportable tests?

Hi Simon, > > How about > > 1) change the default for --create-testdir so that it includes unportable > > or > > longrunning tests only for the modules explicitly listed on the command > > line, not for the dependencies? > > 2) adding options --without-unportable-tests etc.? > > > >

ttyname_r on Mac

Note: this problem doesn't cause my any immediate problem, but it breaks the daily build on Mac. I don't think it is important to work on fixing this, but someone finds it interesting. There appears to be a ttyname_r signature bug, see: http://autobuild.josefsson.org/gnulib/log-201004250717723515

Re: portable openpty, forkpty

Bruno Haible writes: > I'm committing this, clarifying that the master file descriptor needs to be > closed before the slave file descriptor: Thanks. I may get access to a newer Mac relatively soon, and I'll test it on that machine as well. /Simon

Re: portable openpty, forkpty

Hi Simon, > The stand-alone program below reproduces the Mac OS X bug: > > espresso:~ jas$ cc -Wall -o test-openpty test-openpty.c > espresso:~ jas$ ./test-openpty master > Closing master: 0 > espresso:~ jas$ ./test-openpty slave > ^C > espresso:~ jas$ ./test-openpty > ^C^Z^C^Z^C^Z^C^Z > Bruno,

Re: strnlen on OSX & C++

Simon Josefsson wrote: > > This is the same issue as with fchownat before, and this patch makes > > the compile and run successful: > > Is the patch the right thing? No, the proposed patch is not optimal: On platforms where the function simply does not exist (such as MacOS X or Solaris), it cause

Re: portable openpty, forkpty

Bruno Haible writes: o> Hi Simon, > >> Now the test-openpty self-test hangs on my Mac OS X machine, and the >> process doesn't even go away from the process table after a 'kill -9'. >> Here is a gdb session, which suggests it hangs when shutting down the >> process: > > Strange. This program now

Re: test-string-c++.o:(.data+0x0): undefined reference to `rpl_memchr'

Bruno Haible writes: > Simon Josefsson wrote: >> Adding AC_PROG_CXX to gltests/configure.ac solves >> the problem. I see that ansi-c++-opt.m4 may be attempting to do setup a >> C++ compiler, but it doesn't seem to work when cross-compiling. > > Well spotted! This should fix it: It works. Thank