Continuing with today's HEAD's results: 7506 expected passes 235 expected failures 0 unexpected passes 9 unexpected failures More failures than Friday, with fewer tests run (no profiling). But what's the actual difference? We have our old acquaintances T3064(normal) T5084(normal) dph-diophantine-opt(normal,threaded1,threaded2) dph-words-opt(normal) T3016(profasm) obviously wasn't run, without profiling support. New are: ds022(normal) an overlapping pattern warning used to report the first overlapping patterns, now the last; the expected stderr output has been updated meanwhile, but my testsuite was started before that. 3307(normal) environment001(normal) These two failed with (mutatis mutandis) Actual stderr output differs from expected: --- /dev/null 2011-05-15 19:13:45.604004218 +0200 +++ ./lib/IO/environment001.run.stderr.normalised 2011-05-16 12:33:18.000000000 +0200 @@ -0,0 +1 @@ +/bin/sh: Zeile 0: echo: Schreibfehler: Datenübergabe unterbrochen (broken pipe). *** unexpected failure for environment001(normal) I think it's the testsuite driver interacting badly with my OS, because running `make TEST=XY' in the testsuite directory consistently fails with a broken pipe, but running the tests from the .../lib/IO directory, either by manually issuing the commands from the command line or by running `make XY-test', consistently succeeds (including the production of the correct output). Altogether, things still look good.