User tests: Successful: Unsuccessful:
A small regression was introduced into JInstaller
via the loadAdapter()
method which required all adapter classes to have a JInstallerAdapter
prefix. One of the changes in this PR addresses that by restoring the behavior from the foreach loop in JAdapter::loadAllAdapters()
to allow alternate class names and attempting to manually load the file. The manual load code is deprecated for 4.0 and the caller should either perform this function themselves or make their code autoloadable.
Next, the JInstallerAdapter
class which extends from JAdapterInstance
had a more lenient constructor by not typehinting the database and option params. This is changed to match the parent class' typehinting. It should be B/C to 3.3 as the extension adapters previously all extended from JAdapterInstance
and the JInstallerAdapter
class is a new intermediary class in that chain.
Lastly, the JInstaller
constructor did not accept any of the params that its parent, JAdapter
, and instead always injected its forced params into the class. The constructor now allows the params to be injected and will fall back to the previous default values if custom values are not injected.
Labels |
Added:
?
|
Category | ⇒ | Installation Libraries |
Milestone |
Added: |
@digitalgarage No this doesn't address #5977.
Aside from strict error in older version of my JInstaller override, the code works well and I can confirm it fixes all the issues I had previously.
As it looks like that J2.5 and J3.3 versions of the JInstaller class did override constructor, it is probably best to keep the interface as it was before to avoid strict error on extended classes.
Any news on getting this one in? Another test needed?
Status | Pending | ⇒ | Closed |
Closed_Date | 0000-00-00 00:00:00 | ⇒ | 2015-02-16 14:07:59 |
Milestone |
Added: |
Hi Michael, was this for fixing #5977 ?
I just applied applied patch and then tried installing a component with multiple calls to JInstaller->install without calling new JInstaller each time and got the same error.
I have attached screenshot.
This is the only testing I did though so maybe it was to correct something different?
Eric.