|
Posted by Herbert Poul  |
|
i think you are right.. it is currently not behaving correctly on error.. but in my opinion it should still mark it as complete.. but warn the user about it.. and also tell the user which statements to run by hand (since these are mostly alter statements there is no transaction .. hence no need to run all statements from scratch)
i have created an issue for this problem: http://code.google.com/p/sct-project/issues/detail?id=138 Hey, we have Signatures !!! Great, isn't it ? ;) |
|
|
Posted by David Abrahams  |
|
|
I can't imagine what could possibly justify labelling an incomplete operation as complete. Could you explain, please? Thanks. |
|
|
Posted by Herbert Poul |
|
|
if it would be marked as complete when it fails.. there is no way to ever mark it as complete.. if you run it by hand and then again run a syncdb it will fail because all changes were already applied by the user..
there is no reasonable way to detect if all changes were successfully completed by the user imho.. the changelog was never meant to be a complete schema evolution framework or something like that.. just a small helper to keep track of what changes are required for an installation after an update .. and i think it handles this job good enough .. Hey, we have Signatures !!! Great, isn't it ? ;) |
