[Adium-devl] Caution: Merge Left

[Colin Barrett]

How much testing have you done of 18245? all the other revisions look  
fine to me.

I was going to suggest waiting for svn 1.5 to merge project  
parallelization (mmm, cherry picking), but looking at their issue  
tracker they don't seem to have entered "release mode" for it yet, so  
I think it's a bit of a ways off.

HOLD [18096] (remove version.plist). We don't want to break pre- 
sparkle clients for a major release like this! Alternately, restore  
it. Alternately still, if this has been done, update the website to  
reflect it :)

Question: Are the headers for Adium.framework and  
AIUtilities.framework still included? If not, those should be turned  
on for release builds, IMO. Shipping frameworks with headers on for  
frameworks we intend people to use seems like a Good Thing to do. The  
cost in terms of binary size may outweigh this Good Samaritan benefit  
though.

Just as a further test, it would be interesting to see how much time  
is saved on a build where no changes have been made -- i suspect a lot  
of the time spent doing incremental builds is done in the things  
you've removed. If so, we could see a great speedup in workflow,  
especially on slower macs.

Great job, Peter :)

-Colin

On Nov 17, 2006, at 5:08 PM, Peter Hosey wrote:

> I have a couple of big merges to do that I wanted to run by you guys.
>
> == ProjectParallelization ==
>
> http://trac.adiumx.com/wiki/ProjectParallelization
>
> My current plan is to merge everything except the localization  
> changes, putting that off until 1.0 comes out.
>
> Can you guys think of anything else that should be held?
>
> == Memory usage ==
>
> We made a lot of headway on Adium's memory usage when loading the  
> hypahrat log. The relevant changesets are:
>
>  * 18233 (Bigger read size)
>  * 18236 (Use one NSData instead of many)
>  * 18237 (Cleaner loop than 18236)
>  * 18243 (LMX r72 for 18245)
>  * 18245 (Create AIContentObject and throw away AIXMLElement tree  
> after every <message>, rather than when the five-message requirement  
> is satisfied; reduces peak memory usage to only that of the largest  
> message of the five)
>  * 18246 (Header for 18245)
>
> The argument in favor of merging this to 1.0 is that it is a  
> usability concern: the entire Mac becomes unusably slow (because of  
> paging) when a log like the hypahrat log is loaded.
>
> The argument against merging it to 1.0 is that it is exceedingly  
> unlikely that anybody will every encounter such a long. The large  
> messages of the log were contrived by Kent for the specific purpose  
> of stress-testing*; normal usage should never encounter a 1.2 MiB  
> message.
>
> I'm inclined to merge it. On my Mac, a mere 2000 elements (BRs or  
> style changes) is enough to cause a noticeable delay; five large  
> messages could well achieve that. And the bar is lower on less- 
> powerful Macs, such as I imagine the average Adium user runs. It's a  
> *lot* faster to load such logs on trunk that in b15.
>
> * Something else, not message context.
> ________________________________
> \ Peter Hosey / prh at boredzo.org
> PGP public key ID: 7AB26BAD (since 2006-01-01)
>
> _______________________________________________
> Adium-devl mailing list
> Adium-devl at adiumx.com
> http://adiumx.com/mailman/listinfo/adium-devl_adiumx.com