Since there were no objections, I committed this change. tm_gmtoff on
has the same semantics now on Windows as on Unix.
Branko Äibej wrote:
Subversion got bitten by the APR time functions on Windows again. The
culprit, as usual, was incorrect semantics of the tm_gmtoff field in
apr_time_exp_t in the presence of DST. I decided that enough was
enough, and it was time to do something about it.
It turned out that none of the tests in testtime.c tickled that
particular problem, so I added a test case (see patch below). The test
passed on Unix with and withoug DST, and failed on Windows with DST
enabled; again, as expected.
I tracked down the provblem to the way tm_gmtoff was computed from the
system's timezone info. Reading the docs for GetTimezoneInformation
suggested that we should be using the StandardBias and DaylightBias
fields compute tm_gmtoff, so that's what I did. The test now passes on
Win2k (and presumably NT and XP) both with and without DST, and the
relevant Subversion test passes, too.
Now, before I commit this patch, I'd like to ask people to test it on
Win95/98/Me (and possibly WinCE). I don't have access to any of these
platforms, and the docs suggest there may be some difference on the
Win9x series at least.
Thanks,
Brane
--
Brane Äibej <[EMAIL PROTECTED]> http://www.xbc.nu/brane/
