Juiceman skrev:
>>>> 4) Would it be a good idea (read: worth the time) to make the installer
>>>> log its progress to a logfile in the system's temporary folder for
>>>> debugging/troubleshooting reasons?
>>>>
>>>>         
>>> I don't know that users could find the file if needed.  I would log it
>>> in the Freenet install folder if it can be, fall back to the folder
>>> the installer is running and fall back to the system temporary folder
>>> if the others aren't creatable\writable.
>>>
>>>       
>> But if the installer actually succeeds, the user most likely doesn't
>> want/need the log file - so keeping it after the installation would
>> probably be a waste. We won't even know the install folder when we need
>> to log the first things btw. On the other hand, admins doing silent
>> installs would probably be happy with a logfile even if the install
>> succeeds (but they would be doing it via the command-line, where they
>> could also specify a log location).
>>
>> If logged to the temp folder, the installer will delete it during its
>> cleanup on an "expected" exit (= it will only remain there if something
>> went wrong). But on the other hand, as you mention, the user won't
>> realize where he needs to look in the event of a fatal
>> crash/freeze/bug/something.
>>
>> Hmm.
>>
>>     
>
> So create the log file in the .\ folder (the same folder as installer
> is run from, assuming write privledges), always deleting upon success.
>  Also create a log file in the system temp file, deleting on success?
>   

That could be a solution. Anyone see a problem with this?

>>>> 6) update.cmd was *not* created with concurrent installations in mind
>>>> (read: it won't work), and certainly won't be compatible with the
>>>> current installer and mine at the same time. The best solution would
>>>> probably be to branch update.cmd for my installer. Juiceman, how are you
>>>> feeling about it (being the current maintainer)?
>>>>
>>>>
>>>>         
>>> I haven't worked on it in a while.  Why would it not work?  It stays
>>> in it's own directory and uses relative folders ..\bin etc etc.
>>> Unless you are wanting it to update all installations at once (bad
>>> idea imo) it just calls stop.cmd does it's thing and calls start.cmd.
>>> All agnostic, right? or am I missing something.  I assume the new
>>> installer will create a start/stop.cmd with the unique nodes' service
>>> name?
>>>
>>> I would be happy to make changes as needed.
>>>
>>>       
>> Coolio. I agree that it should only update its own installation. Yes,
>> new start.exe and stop.exe bins will be included. They will be generic
>> though, as they will read the needed ID info from a file (which you can
>> adapt for update.cmd as well :)):
>>     
>
> Wait, .exe?
>   

Yea. To start/stop services under Vista, you need to UAC elevate, which 
cannot be done from a batch script :-/. See 
http://en.wikipedia.org/wiki/User_Account_Control#Requesting_elevation. 
Basically we either have to include an .exe manifest (my current 
solution as it seems like the best approach) or elevate access by 
"runas"'ing ourselves.

>> However, there have been some Vista bug reports regarding various things
>> failing because of missing privileges (UAC :-/). So the "cacls" stuff
>> might not even work in a .cmd file on Vista, meaning we might have to
>> create wrappers for that too (as start.exe/stop.exe which will
>> self-elevate on Vista). I still need a Vista test puppy to test that
>> kind of stuff though.
>>     
>
> I'll have to read up on the cacls stuff, I didn't put that in Toad or
> Nextgens did.
>
> Yuck.  I don't have a Vista machine to test on, either.
>   

It's basically chmod for Windows. If we are lucky, the user just gets an 
UAC promt. If we are unlucky, it fails like "net start" and "net stop" :-/.


- Zero3
_______________________________________________
Devl mailing list
[email protected]
http://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl

Reply via email to