Jim Meyering <jim <at> meyering.net> writes: > In that case, the configure-time check for is-an-ELF-system > must be failing on cygwin:
No: configure:43745: checking whether this is an ELF system configure:43761: result: no But there might be some missing holes where 'make check' tries to build stdbuf without regards to the result of the configure results. > > > > I'm now running 'make -k check', and will try to spot any obvious problems. So far, I'm seeing a suspicious failure: tests/misc/factor.log: factor (GNU coreutils) 7.4.121-ea5b7 ... z... factor: test z: stderr mismatch, comparing z.E (actual) and z.1 (expected) *** z.E Fri Aug 14 10:36:48 2009 --- z.1 Fri Aug 14 10:36:48 2009 *************** *** 1,2 **** ! factor: unknown option -- 1 Try `factor --help' for more information. --- 1,2 ---- ! factor: invalid option -- 1 Try `factor --help' for more information. cont... I'm not sure whether that might be related to the getopt vs. getopt-gnu change though (I've been building incrementally, and may have picked up some state where cygwin's getopt snuck into usage for a while; I'm going to repeat with a pristine bootstrap to see if the problem still persists). The test is still running. Meanwhile, there are a number of failures probably due to the fact that cygwin Admin users can read and delete files that are otherwise unreadable from POSIX permission standpoint (such as rm/rm1). But this is no different than in the past; I've been meaning to find time to write some sort of test filter that checks for superuser privileges to skip these sorts of tests, and I can also try and figure out how to repeat the testsuite run under a different user without Admin rights. Again, don't hold up the release waiting for me (I've had a lot on my plate lately on other fronts, and I maintain the cygwin port of coreutils so whatever doesn't make it upstream will still make it by the time I complete the cygwin packaging). -- Eric Blake