Re: CVS Time Zone Bug

[Derek Robert Price]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Chris Bohn wrote:

> OK; while it was a pain to get through, here are the results:
>
> The 1.12.9 source distribution doesn't build under Windows.  After
fixing a few of the errors, I looked at the source tree and found you
applied many fixes submitted by Conrad so that it would build under
Windows.  After applying a few of them as I saw the updates in the
head of the project and continuing to hit more, I eventually just
copied the head of the ccvs project's Windows-NT directory to my
1.12.9 source tree (which resulted in the builds having version
1.12.9.1).  With that, and updated .mak files for some of the
directories, it finally built.  I wasted a lot of time with this.  I
think we should come up with a way of indicating when the "released"
source doesn't build on a particular platform, and when the
appropriate fixes have been made, it would be nice for a point release
to be made so the source can be built on all platforms.  Both would
have saved me a lot of time, and just indicating the source wouldn't
build under Windows would have also.  That is a separate issue, but it
really would be helpful.

Sorry about that.  There hasn't been anyone actively working on our
Windows port in awhile.  Due to this, the pattern we have been
following is, unfortunately: 1)  Release CVS,  2)  Get lots of bug
reports about Windows build problems, 3)  Fix Windows build problems
in dev, 4)  Fix unassociated problems, thereby breaking Windows build
again, 5)  Release CVS.

Luckily for our Windows uses, though, we jut picked up Conrad Pino on
Windows and I just started trying to notify him about impending
releases, so maybe this will change.

> After applying the patch you gave, it initially wouldn't build.  I
had to change all 'bool' references (standard only in C++) to 'BOOL'.

Not true.  `bool' is standard in C99.  The correct fix for this is to
#include <stdbool.h> in the JmgStat.c file.  I've committed this.

> Once that was fixed, I got errors indicating "Failed to retrieve
modification time for xxx".  I had to change filesubr.c from
>
> if (GetUTCFileModTime (file, &sb->st_mtime))
> to
> if (!GetUTCFileModTime (file, &sb->st_mtime))

Thanks.  I've al=pplied this change and committed the patch.

> (just adding the '!' since true is returned on success).  After
doing that, it built and seems to fix the problem.  I did notice that
my 1.10 client in some cases puts the cvs\entries time off by an hour,
so some files are still marked incorrectly as having changed (cvs stat
fixes the time), but hopefully this is just a one time change since my
understanding is that changing to/from DST will no longer make every
file listed as having been changed (the way 1.10 does).

Yes, this will hopefully be a one time change and I would not expect
compatibility with old CVS clients accessing the same workspace.
Since old versions of CVS were incorrectly retrieving mod times, they
were also writing them incorrectly to the Entries file.

> When do you expect to release a real build for this?  I'd like to
roll out the real version and not a custom version.

I just got my head out of a contract, so I'm currently trying to get
through my back log of CVS patches I wanted in the next release, but
it should be soon.

> thanks

Thank you!

Cheers,

Derek
- --
                *8^)

Email: derek@ximbiot.com

Get CVS support at <http://ximbiot.com>!
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFBPgNZLD1OTBfyMaQRAhy4AKDKI3i2uOM5rLLYGSCVwfMYSIpj2ACg3hWr
ucjbTX3R5C95nogLiNqd+LA=
=ZulW
-----END PGP SIGNATURE-----