
It appears that someone reinstalled chimera on ryle and that blew away all of the development only stuff in the Chimera installation tree. Please don't ever do that again. Thanks, Greg

Hi Greg, I think the Chimera build system should not install Chimera as each piece is built. The trouble with that is if the build fails half way through, then the installed Chimera is crippled. I don't want that to happen even on our development machines. I think Chimera should build into a temporary build directory, and when the build succeeds, a "make install" puts it in the installed location. This is an easy change to our current system, simply pointing the Chimera build to a temporary directory instead of the install directory. Tom

On Wed, 28 Aug 2002, Thomas Goddard wrote:
I think the Chimera build system should not install Chimera as each piece is built. The trouble with that is if the build fails half way through, then the installed Chimera is crippled. I don't want that to happen even on our development machines. I think Chimera should build into a temporary build directory, and when the build succeeds, a "make install" puts it in the installed location. This is an easy change to our current system, simply pointing the Chimera build to a temporary directory instead of the install directory.
Actually we have such a two-stage system for the PCs, it's called ryle. I'm thinking of modifying the installer so it won't run on ryle. As for the other systems, we already have the equivalent of a temporary build directory, the source/obj directory where the code is built. Chimera can use that version very easily (with --pypath) without overwriting the installed version and "make install" behaves as described. Perhaps a little more restraint on the part of developers to wait until after testing their changes to install them is needed (myself included). Greg

On Tuesday, August 27, 2002, at 07:39 PM, Greg Couch wrote:
It appears that someone reinstalled chimera on ryle and that blew away all of the development only stuff in the Chimera installation tree. Please don't ever do that again.
Well, that was me. I needed a Windows system to test a problem in the 1516 release and arnauld and the adjacent Windows boxes were in use for collaboratory testing. Ryle only had the 1442 release. Clearly, I should have chosen the "move old installation" instead of "delete old installation" installation option. That was retarded on my part. It might have been prudent on your part, once a significant amount of time had been invested in changes, to move the modified version of Chimera to another location. It would be misleading anyway to leave the modified version as the "1442 release". [Nonetheless, I take the blame here -- just a suggestion]. --Eric

Eric Pettersen wrote:
On Tuesday, August 27, 2002, at 07:39 PM, Greg Couch wrote:
It appears that someone reinstalled chimera on ryle and that blew away all of the development only stuff in the Chimera installation tree. Please don't ever do that again.
Well, that was me. I needed a Windows system to test a problem in the 1516 release and arnauld and the adjacent Windows boxes were in use for collaboratory testing. Ryle only had the 1442 release.
Clearly, I should have chosen the "move old installation" instead of "delete old installation" installation option. That was retarded on my part.
It might have been prudent on your part, once a significant amount of time had been invested in changes, to move the modified version of Chimera to another location. It would be misleading anyway to leave the modified version as the "1442 release". [Nonetheless, I take the blame here -- just a suggestion].
Since much (most) of the Chimera testing does not actually require hardware graphics, I strongly suggest you guys just make a test Windows 2000 VMware partition that can be used to install/test chimera. The really, really nice thing about VMware is you can use a 'rollback segment' for the disk writes- so that at the end of testing, when you shut down the machine, you can choose to commit your changes to the disk repository or dump them. If you dump them then when the machine reboots it's as if it was never touched (virgin install). Or you can commit them and fork the disk repository, keeping the old and the new. Dave

On Wed, 28 Aug 2002, Eric Pettersen wrote:
Well, that was me. I needed a Windows system to test a problem in the 1516 release and arnauld and the adjacent Windows boxes were in use for collaboratory testing. Ryle only had the 1442 release.
Clearly, I should have chosen the "move old installation" instead of "delete old installation" installation option. That was retarded on my part.
It might have been prudent on your part, once a significant amount of time had been invested in changes, to move the modified version of Chimera to another location. It would be misleading anyway to leave the modified version as the "1442 release". [Nonetheless, I take the blame here -- just a suggestion].
So now I know what happened. Release 1442 was installed on ryle in a temporary directory (by Eric :-)) to compare against 1516. Both were on ryle, but the shortcuts were changed to 1442 because it was installed last. So a reinstall wasn't needed. Too late now. Maybe Eric could add ryle to the systems that are backed up -- just a suggestion. Greg
participants (5)
-
David E. Konerding
-
Eric Pettersen
-
Greg Couch
-
Greg Couch
-
Thomas Goddard