Re: [Pharo-project] How to correctly remove a package for everywhere?
Seems that removing a package does not imply removing the included classes?Maybe this should be changed? With RPackage is understandable, since its just a layer on top of the system.You have to explicitly register/deregister packages. Fernando On Wed, Jun 20, 2012 at 10:24 PM, Mariano Martinez Peck <marianopeck@gmail.com> wrote:
well, I don't know what the hell happened, but a workaround I think it is
DataStream removeFromSystem. RPackageOrganizer  default unregisterPackageNamed: #'System-Object Storage'.
On Wed, Jun 20, 2012 at 10:19 PM, Mariano Martinez Peck <marianopeck@gmail.com<mailto:marianopeck@gmail.com>> wrote: So removing 'System-Object Storage' didn't work. I did:
ScriptLoader new deletePackageNamed: 'System-Object Storage'.
Now the MCWorkingCopy is removed (I don't see the package in Monticello Browser), but I can still browse the class DataStream, and the package is available from Nautilus.
Now if I do:
RPackageOrganizer  default unregisterPackageNamed: #'System-Object Storage'
I don't see the package in Nautilus, but the class "DataStream" is still alive. I can do "DataStream name -> #DataStream" ..et...when I browse it with Nautilus I don't see the class but the category _UnpackagedPackage..
so...how the hell can I remove this package and class from everywhere in the system?
thanks
-- Mariano http://marianopeck.wordpress.com
-- Mariano http://marianopeck.wordpress.com
In Monitcello days (and browser), there is Delete working package, and Unload package. I suspect deletePackageNamed: is all about removing the idea of the package, but not the code (as you found) - just making the code part of the image and not a package. You want something like unloadPackage: or #unloadPackage to remove the code (and package) from the system. Assuming that is still there, that is (I don't have a version 2.0 yet). Wouldn't your drive to package everything make you want to guard against the deletePackageNamed: behaviour? Only delete it if there is no code associated to the package, otherwise complain somehow. -Chris
participants (2)
-
Chris Cunningham -
Fernando Olivero