On 4/20/05, Alexandre Julliard julliard@winehq.org wrote:
Modified files: server : registry.c
Log message: Load system.reg and userdef.reg at server init time.
Hi Alexandre,
A while back I sent in a patch that restricted RegCreateKey (and it's internal implementation) from creating keys in HKLM and HKU as stated by msdn and the tests, and you replied:
I don't think there's any point in replicating that behavior. I doubt there's any such check on Windows either, it's most likely just a side effect of the way the registry files are "mounted" at different places in the tree. Once we have a proper implementation of RegLoadKey/RegUnloadKey we should get the same behavior automatically.
We now have an implementation of RegLoadKey and I was wondering if it's time to switch from CreateKey to LoadKey in the registry initialization. If not, what remains to be done? Aside from the reg init topic, is there any other change needed to the registry subsection? I guess what I'm asking for is a registry TODO if you have one in mind.
Thanks, James Hawkins
James Hawkins truiken@gmail.com writes:
I don't think there's any point in replicating that behavior. I doubt there's any such check on Windows either, it's most likely just a side effect of the way the registry files are "mounted" at different places in the tree. Once we have a proper implementation of RegLoadKey/RegUnloadKey we should get the same behavior automatically.
We now have an implementation of RegLoadKey and I was wondering if it's time to switch from CreateKey to LoadKey in the registry initialization.
Registry initialization is already doing a LoadKey, there isn't really anything to change here. We still don't have a proper implementation of registry mount points though, but that's not really necessary for the init code, it's only for apps that try to load extra registry files.
On 20 Apr 2005 18:57:25 +0200, Alexandre Julliard julliard@winehq.org wrote:
Registry initialization is already doing a LoadKey, there isn't really anything to change here. We still don't have a proper implementation of registry mount points though, but that's not really necessary for the init code, it's only for apps that try to load extra registry files.
What would a proper implementation of registry mount points require, or what is the current implementation doing wrong? Do we want our implementation to be more in line with windows'?
James Hawkins truiken@gmail.com writes:
What would a proper implementation of registry mount points require, or what is the current implementation doing wrong? Do we want our implementation to be more in line with windows'?
I think the current implementation is mostly behaving correctly, except for details like the RegCreateKey difference you noticed. Very few apps use these APIs anyway, so I'm not sure there is much need to fix the border cases, unless we find something that depends on them.
-
Alexandre Julliard -
James Hawkins