Re: Bug (old?) in ZipEntry?

[Martin Egholm Nielsen <martin at egholm-nielsen dot dk>]

> > > I think I have found the problem. The "fromzip.zip" archive
> > > contains an additional timestamp entry for "ZipDates.java":
> > >
> > > Current Location part 1 offset 43
> > >   extra field 0x5455 (universal time), 4 header and 9 data bytes:
> > >   03 90 5d 21 43 f6 5d 21 43                      ..]!C.]!C
> > >   extra field 0x7855 (Unix UID/GID), 4 header and 4 data bytes:
> > >   20 02 11 29                                      ..)
> > >
> > > The modification timestamp is "1126260112" which is *seconds*
> > > since *the standard UNIX epoch* (i.e., it is *not* a dostime).
>
> [...]
>
> > > Our ZipEntry code (both in libgcj/classpath) treats it as a
> > > dostime, leading to the error that you have noticed.
>
> This part is wrong - I misread the code.
>
> > > Please file a bug report against classpath, including a link to
> > > this message, if possible.
> >
> > Hereby done!
>
> If I patch GCJ's java/util/zip/ZipEntry.java as shown later, I get:
>
>   ~/tmp/men > mygij ZipDates fromzip.zip
>   ZipDates.java   date: 1126260112000     Fri Sep 09 15:31:52 GMT+05:30 2005
>
>   ~/tmp/men > java ZipDates fromzip.zip
>   ZipDates.java   date: 1126247512000     Fri Sep 09 12:01:52 IST 2005
>
> Note that GCJ still seems off by 3.5 hours. 

> However, according to the
> docs, the given timestamp is seconds from epoch *UTC*, which *does*
> translate to the time shown by GCJ. To confuse things further, the ZIP's
> "normal" entry for "ZipDates.java" (not the 0x5455 extension entry
> shown earlier) says it is the time reported by the JDK. (What is your
> timezone? Is it GMT+02:00?)
Yes! So the 3.5h offset is the difference between you and me.
I created the file 12:01:52 GMT+02:00 (so it's really 10:01 UTC), the 
time JDK reports for you, however, in your timezone (so it thinks it has 
time 06:31 UTC - JDK error).
So, actually GCJ/Classpath does it correctly, since 15:31:52 GMT+05:30 
is 10:31 UTC - the actual time of the file.

(Messy written - I know)

> So either the JDK is showing only the "normal" entry without taking
> the extension entry into account or the reference document that I
> cited earlier has it wrong (unlikely). Either ways, your original
> archiver program (the one that created the "fromzip.zip" archive)
> is lying about the modification time in one of the entries.
Yes, I think my archiver puts the time in plus the timezone offset...

> The patch I was talking of was:
I'm doing a new libgcj as we speak - looking forward to try it!

Thanks Ranjit!

// Martin

> - ------------------------------- 8< -------------------------------
> Index: java/util/zip/ZipEntry.java
> ===================================================================
> - --- java/util/zip/ZipEntry.java 2005-09-13 13:40:41.000000000 +0530
> +++ java/util/zip/ZipEntry.java 2005-09-13 15:43:40.000000000 +0530
> @@ -168,5 +168,5 @@ public class ZipEntry implements ZipCons
>      synchronized (cal)
>        {
> - -       cal.setTime(new Date(time));
> +       cal.setTimeInMillis(time);
>         dostime = (cal.get(Calendar.YEAR) - 1980 & 0x7f) << 25
>           | (cal.get(Calendar.MONTH) + 1) << 21
> @@ -337,5 +337,5 @@ public class ZipEntry implements ZipCons
>                             | (extra[pos+3] & 0xff) << 16
>                             | (extra[pos+4] & 0xff) << 24);
> - -                   setTime(time);
> +                   setTime(time*1000);
>                   }
>               }
> - ------------------------------- 8< -------------------------------
>
> Ranjit.
>
> - --
> Ranjit Mathew      Email: rmathew AT gmail DOT com
>
> Bangalore, INDIA.    Web: http://ranjitmathew.hostingzero.com/
>
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.2 (MingW32)
> Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
>
> iD8DBQFDJq8rYb1hx2wRS48RAsgEAJwOuoDs1RVvhj/GO7DRY6LixsPHaACeN+GW
> E2iucXRl4mRLmTiHAA9c+zg=
> =LpSq
> -----END PGP SIGNATURE-----