Hello Groups, I am asking for your help on another problem in XEmacs on Windows NT native: fstat and stat seem wrong on Windows NT. XEmacs tries to work around the stat problem by implementing its own version. It does however use the equally buggy fstat version of Windows NT. This creates the issue that .el files appear newer than their .elc files (created while DST was on) after DST is switches off. Inside DEFUN ("load-internal", Fload_internal, 1, 6, 0, /* in lread.c the .elc file is fstat'ed with NT's native implementation while the .el file is stat'ed with XEmacs's stat implementation. perl (This is perl, version 5.005_03 built for MSWin32-x86) gives the same results as the Windows NT compatibility function stat. Here's what I see: My computer is in the MET-1MDT timezone. I have not explicitely set env. var. TZ in the experiments below, but that did not help when I tried. With system date "nov 8 1999" I create a file called "winter". I switch the system date to "oct 8 1999" and create "summer". Now I get different time_t values for their modification time both in perl and Windows NT functions fstat/stat, depending on the current date of the computer: With computer date "oct 8 1999": d:\Users\AichnerAd\tz\summer st_mtime: 939409806 gmtime Fri Oct 8 19:10:06 1999 localtime Fri Oct 8 21:10:06 1999 d:\Users\AichnerAd\tz\winter st_mtime: 942095185 gmtime Mon Nov 8 21:06:25 1999 localtime Mon Nov 8 22:06:25 1999 With computer date "nov 8 1999": d:\Users\AichnerAd\tz\summer st_mtime: 939406206 gmtime Fri Oct 8 18:10:06 1999 localtime Fri Oct 8 20:10:06 1999 d:\Users\AichnerAd\tz\winter st_mtime: 942091585 gmtime Mon Nov 8 20:06:25 1999 localtime Mon Nov 8 21:06:25 1999 I am happy with the distances between GMT (=UTC) and localtime, which are 2 hours in summer, 1 hour in winter (for MidEuropeanTime). However, the supposedly absolute modification time of a file changes with the current computer date setting (see st_mtime: ...) above. Please confirm that this is actually a Windows NT bug! FileTimeToLocalFileTime has this bothering remark, that a FILETIME will be adjusted according to the current DST status, not according to the file's time. Yuck! Could the knowledgable please comment? Currently I suppose the fix is to also encapsulate fstat in XEmacs Windows NT native. Any suggestions? Best regards, Adrian