Re: [PATCH try2 2/2] winemenubuilder: Wait for application icons to be created instead of adding a RunOnce entry that may never run.
On Wednesday 27 June 2007, Misha Koshelev wrote: + /* Now watch the directory for our path to be created */ + if ((*handles = FindFirstChangeNotificationW(dir, subtree, FILE_NOTIFY_CHANGE_FILE_NAME)) == + INVALID_HANDLE_VALUE) I'm not sure if file change notifications are the right solution here but since they are only supported on Linux you should at least keep the run once key as a fallback strategy. -Hans
On Wed, 2007-06-27 at 12:37 +0200, Hans Leidekker wrote:
On Wednesday 27 June 2007, Misha Koshelev wrote:
+ /* Now watch the directory for our path to be created */ + if ((*handles = FindFirstChangeNotificationW(dir, subtree, FILE_NOTIFY_CHANGE_FILE_NAME)) == + INVALID_HANDLE_VALUE)
I'm not sure if file change notifications are the right solution here but since they are only supported on Linux you should at least keep the run once key as a fallback strategy.
-Hans
Hmm... I was not aware of this (that means that make change.ok in dlls/kernel32/tests fails on non-Linux systems?). What happens in these other non-Linux systems... does FindFirstChangeNotification return INVALID_HANDLE_VALUE or succeed and just never satisfy the wait condition? The good thing about the __wine_make_system_process wait we are also waiting for is that it will get satisfied when all user processes have closed and create links then as well. This is actually sufficient in itself for some installers (Vector NTI), but not others that launch apps while installing (iTunes). Perhaps I can add a long timeout to the wait to account for these other non-Linux platforms... Anyhow, please let me know about the behavior of FindFirstChangeNotification on these other systems or where I can find this documented... Misha
Misha Koshelev <mk144210(a)bcm.edu> writes:
Hmm... I was not aware of this (that means that make change.ok in dlls/kernel32/tests fails on non-Linux systems?). What happens in these other non-Linux systems... does FindFirstChangeNotification return INVALID_HANDLE_VALUE or succeed and just never satisfy the wait condition?
It will never satisfy the wait. That can happen on older Linux kernels too, or on network file systems. You can't rely on change notifications, they are just a hint.
The good thing about the __wine_make_system_process wait we are also waiting for is that it will get satisfied when all user processes have closed and create links then as well. This is actually sufficient in itself for some installers (Vector NTI), but not others that launch apps while installing (iTunes). Perhaps I can add a long timeout to the wait to account for these other non-Linux platforms...
A better approach would probably be to simply wait for the parent process to exit. -- Alexandre Julliard julliard(a)winehq.org
From: Alexandre Julliard [mailto:julliard(a)winehq.org] Sent: Wed 6/27/2007 8:42 AM To: Koshelev, Misha Vladislavo Cc: Hans Leidekker; wine-devel(a)winehq.org Subject: Re: [PATCH try2 2/2] winemenubuilder: Wait for application icons to be created instead of adding a >RunOnce entry that may never run. Misha Koshelev <mk144210(a)bcm.edu> writes:
Hmm... I was not aware of this (that means that make change.ok in dlls/kernel32/tests fails on non-Linux systems?). What happens in these other non-Linux systems... does FindFirstChangeNotification return INVALID_HANDLE_VALUE or succeed and just never satisfy the wait condition?
It will never satisfy the wait. That can happen on older Linux kernels too, or on network file systems. You can't rely on change notifications, they are just a hint.
The good thing about the __wine_make_system_process wait we are also waiting for is that it will get satisfied when all user processes have closed and create links then as well. This is actually sufficient in itself for some installers (Vector NTI), but not others that launch apps while installing (iTunes). Perhaps I can add a long timeout to the wait to account for these other non-Linux platforms...
A better approach would probably be to simply wait for the parent process to exit.
Good point. This would probably avoid the __wine_make_system_process call altogether too. Will resubmit again prob sometime tonight. Misha
participants (4)
-
Alexandre Julliard -
Hans Leidekker -
Koshelev, Misha Vladislavo -
Misha Koshelev