Paul Richard Thomas wrote: > This seems to have fixed itself on x86_ia64/FC5 and amd64/Cygwin_NT. > Unless anybody finds a system/OS on which it fails, I propose to > commit the testcase and to close the PR tomorrow night. Fine with me as I cannot reproduce it with 4.3.0 (it fails with 4.2.x and 4.1.3). Tobias