carlton@apollo.HP.COM (Carlton B. Hommel) (08/15/90)
mike@tuvie (Inst.f.Techn.Informatik) describes how the latest patchlevel
for perl fails the following regression test:
($dev,$ino,$mode,$nlink,$uid,$gid,$rdev,$size,$atime,$mtime,$ctime,
$blksize,$blocks) = stat('Op.stat.tmp');
if ($mtime && $mtime != $ctime) {print "ok 4\n";} else {print "not ok 4\n";}
I saw this when I compiled patchlevel 18. I submitted it as a defect, and
the following discussion ensued:
Date engineer baselevel
02/03/90 carlton sr10.3bl45
stat mtime & ctime different for newly created file
[Yes. We don't try to keep the .ctime == .mtime for a newly
created file. (the .mtime is assigned when the file is
created; the .ctime is advanced "just" afterwards as various
attributes are being assigned: protection info, type, etc.)
Do any of the SVID, POSIX or X/OPEN actually stipulate this? ]
[I haven't seen anything in any of the standards that mandate
this behavior. Carlton, is this important for some application? ]
[No, it isn't. -carlton ]
[Nobody promises that they will be the same. ]
To summarize, yes, perl on Domain/OS will occasionally fail this test,
but no, it isn't an indication that anything is wrong with either perl or
Domain/OS.
If it turns out that, indeed, mtime == ctime is mandated, please email me
the citation for the appropriate standards document, and I will reopen
the bug.
Carl Hommel
carlton@apollo.hp.com
"I could write a perl script to do that...."