On Mon, Aug 1, 2016 at 9:27 AM, Jim Meyering wrote: > On Sun, Jul 31, 2016 at 9:35 PM, Jim Meyering wrote: >> On Sun, Jul 31, 2016 at 8:38 PM, Assaf Gordon wrote: >>> Hello, >>> >>>> On Jul 31, 2016, at 21:15, Jim Meyering wrote: >>>> >>>> Here's a new snapshot: >>>> [...] >>>> http://meyering.net/diff/diffutils-3.3.52-e974.tar.xz >>>> >>>> Changes in diffutils since 3.3.50-0353: >>>> >>>> Jim Meyering (2): >>>> tests: colors: fix a portability problem and work around a shell bug >>>> maint: remove gl/lib/reg*.c.diff; no longer needed >>> >>> With this update, no more failures on the tested gnu/linux system, but few failures on others (BSDs, Mac OS X, AIX, OpenSolaris). >>> >>> For BSD/Mac OS X/AIX, it seems the 'color' test uses 'touch --date' which is a gnu extension and is not available. >>> Perhaps that leads to other failures. >>> >>> For OpenSolaris 5.10 (both i86pc and sparc) 'make check' does not complete (perhaps an old shell syntax issue). >>> >>> For GNU Hurd 0.7, the test 'brief-vs-stat-zero-kernel-lies' fails. >> >> Great! Thanks again for all of that quick feedback. >> So far I've only looked at the solaris 5.10 build failure, and wrote >> the attached that I hope will fix it. It upgrades to the >> infrastructure (forgot I hadn't yet done it for diffutils) that is >> used already to good effect in at least grep, sed, coreutils. Would >> you please let me know if this fixes it? > > I've gone ahead and pushed that with a tweak to make envvars check one > more variable name: GREP_OPTIONS. Thanks again for all the testing, Assaf. Would you please see what the hurd test expected? If you rerun it with this patch, the usual testsuite.log file will include details: