On 22/12/15 21:48, Paul Eggert wrote: > Pádraig Brady wrote: >> Oh Solaris has the same issue? > > Sorry, I thought your earlier messages said so; if not then please ignore my > noise about Solaris. > >> So you mean warn, rather than fail or ignore in this case? > > This depends on whether the failure is due to gnulib (in which case we should > fail, and should fix the gnulib bug) or in the C library (in which case we > should just warn). If on Darwin the shell command > > TZ='NZST-12NZDT,M9.5.0,M4.1.0/3' date -r0 > > outputs the string "Thu Jan 1 12:00:00 NZST 1970", then the C library is > clearly buggy. If not, we should investigate further. The above example > assumes > Darwin 'date', not GNU date, so it uses the -r0 option; GNU date would have > -d@0 > instead.
darwin14> TZ='NZST-12NZDT,M9.5.0,M4.1.0/3' /bin/date -r0 Thu Jan 1 12:00:00 NZST 1970 darwin14> TZ='NZST-12NZDT,M9.5.0,M4.1.0/3' /tmp/coreutils-8.24.119-f8a9c/src/date -d@0 Thu Jan 1 12:00:00 NZST 1970 Since we don't replace localtime() a warning is appropriate, and which I've now pushed. thanks, Pádraig.
