sync stdint module from gnulib

2006-07-02 Thread Paul Eggert
I've been hacking on the stdint module of gnulib, and I think it's now good enough for coreutils to use unconditionally, so I installed the following change into coreutils. This should let us simplify a bunch of other configuration code later, but one thing at a time. I've tested this on Solaris

Re: make distcheck fails

2006-07-02 Thread Paul Eggert
Jim Meyering <[EMAIL PROTECTED]> writes: > conftest.c:364: error: negative width in bit-field 'f' > conftest.c:365: error: negative width in bit-field 'g' That's to be expected, since doesn't quite conform to C99 on glibc platforms yet due to glibc bug 2841

Re: tail claims "tail +n" is deprecated

2006-07-02 Thread Alfred M. Szmidt
Breaking backward compatibility is so bad -- I've been bitten many times by autoconf-generated scripts that many GNU packages use and insist on remaking locally (why?) and that could not work because my local autoconf was newer than the version used by the package author. autoconf/a

Re: tail claims "tail +n" is deprecated

2006-07-02 Thread Raphael Manfredi
Quoting Alfred M. Szmidt: : autoconf/automake gets rerun automatically if the timestamps for : configure/Makefile.in is newer than for configure.ac/Makefile.am and : other such files. It often occurs if you are getting something from : CVS. I applaud the intent. I cry in front of the harmful sit

Re: coreutils 5.95 on IRIX 5.3

2006-07-02 Thread Paul Eggert
Georg Schwarz <[EMAIL PROTECTED]> writes: > Here's the full make check output from 5.97. There still seem to be > problems. > ... > rm: cannot remove directory `no-target-dir.tmp/26138': File exists > FAIL: no-target-dir What happens if you do something like this? It should generate more debuggi