[The ChangeLog entry was actually in a subsequent message, but I've joined them here. I have this sinking feeling this would cause some kind of headache, portability or whatever. Maybe I'm being pessimistic, I don't know... -kingdon] Date: Wed, 27 Jan 1999 20:45:05 -0500 (EST) To: bug-cvs@gnu.org Subject: Re: FAIL: client-9 (solaris 2.7) From: Eric Mumpower > > Would it be better to temporarily set the timezone to UTC while performing > > this test? Or is that too hard to do portably in the shell? (I don't know > > if TZ is a portable construct, or if "GMT" is a portable value for it....) > > Sounds like something you might want to investigate ;-). Uli Drepper asserts that TZ is portable because it's necessary for POSIX compliance. He suggests the value "GMT+0" as the most general way of declaring UTC. Thus, I submit this patch as a more demanding version of the aforementioned test. I've tested it on two different hosts, one in PST and one in EST; both passed. 1999-01-27 Eric Mumpower * sanity.sh (client-9): Don't assume the time zone: switch the effective time zone temporarily to GMT. Index: sanity.sh =================================================================== RCS file: /home2/cvsroot/ccvs/src/sanity.sh,v retrieving revision 1.505 diff -r1.505 sanity.sh 18544,18545c18544,18548 < # By specifying the time zone in local time, we don't < # know exactly how that will translate to GMT. --- > # By switching time zones to GMT, we don't have to worry about > # knowing how the local time zone relates to GMT, and so we know > # precisely what hour we're specifying. > TZ_CVS_SAVED=${TZ} > TZ="GMT+0"; export TZ 18552c18555 < Argument [34] Oct 1999 [0-9][0-9]:00:00 -0000 --- > Argument 4 Oct 1999 00:00:00 -0000 18560a18564 > TZ=${TZ_CVS_SAVED}; export TZ