General ports questions

[Scott Haneda]

On Jan 26, 2009, at 5:08 PM, Ryan Schmidt wrote:

> On Jan 26, 2009, at 16:19, Scott Haneda wrote:
>
>> On Jan 26, 2009, at 8:11 AM, Joshua Root wrote:
>>
>>> It's entirely up to you which platforms you want to support in your
>>> ports, and which files are useful to install. Use your judgement.
>>
>> How does one state 10.4 and 10.5 only?
>
>
> The following will exclude users of Mac OS X 10.3:
>
>
> platform darwin 7 {
> 	pre-fetch {
> 		return -code error "${name} requires Mac OS X 10.4 or newer"
> 	}
> }
>
>
> This assumes nobody has MacPorts installed on Mac OS X 10.2 or  
> earlier anymore. MacPorts 1.7.0 doesn't install on Jaguar and  
> earlier anyway.


So next on my list of quandaries is upgrades.
The last port for ASSP was 1.1.0, which is about 4 major releases out  
of date.  I am guessing, no one is using it, those old versions just  
did not work well.

I have to accept there may be someone using 1.1.0 and come along and  
issue a port upgrade, which with the way I have made my port, is going  
to step all over their files.  The entire directory tree is different,  
it will break things for sure.

The trouble is, it sucks to rename the port, and call it assp-new or  
something like that.  What is suggested here?  Is there a way to port  
check for the old version of a port, say, anything 1.1.0 or earlier,  
and just bail out at that point?  I think that would be my best option.

How do upgrades work?  When ASSP first runs, it makes about 5  
directories, these contain logs, and data made by the app once run.  I  
can not mess with these on upgrade.  There are also cfg files as well,  
those need to be left alone.

But, files may get retired, renamed etc, so I can not just put the new  
files in the old place.  Is it ok to copy the old directory, as  
assp.bk or something, install, then pull the important files back into  
the new update?  Or is it best for me to find the diffs in the update,  
and file by file make decisions?

This is for an email server proxy, it is imperative that I test all  
cases, this is not like wireshark breaking on someone.  My sole reason  
for this is for myself, and well as a pretty good handful of email  
admins, most in large ISP installations, and can not have downtime.   
Not all are going to dev stage this, and just go wild with it, and  
while their fault, I want to be as careful as I can.

Thanks for any pointers.
--
Scott