[Adium-devl] Release Plans - 1.3.4/1.4

[Zachary West]

On Sun, May 3, 2009 at 02:54, Peter Hosey <boredzo at adiumx.com> wrote:
> I further propose that we move to DWARF-with-dSYM for our releases,
> including nightlies, alphas, and betas, starting with 1.3.4 and 1.4.
>
> First, I'd like to split out the projects' configurations into xcconfig
> files. One key feature that they have is the ability to #include other
> xcconfig files, which means we can cut out a lot of repetition: Settings
> that we currently repeat in multiple projects, we could have only once in a
> single file.
>
> I'm working on xcconfig-extract again with this goal in mind, but as I write
> this, I'm not sure I'll use it. It may be better to hand-write only a few
> xcconfig files and use those files in every project. An unmodified export of
> the current settings from the projects would not eliminate any repetition at
> all.

This sounds cool.

>
> One way or another, once we have our build settings in xcconfig files, we
> can flip the DWARF-with-dSYM switch[1]. Then, we'll want to modify our
> release-build process to archive the dSYM bundles and publish each archive
> on the website. That should include nightlies as well.
>
> Thereafter, when we get a crash report for a release we have dSYM bundles
> archived for, we'll download the appropriate archive, unzip it, then feed
> the crash log into the Symbolicator[2]. That will symbolicate the crash log
> using the dSYM bundles.
>
> …
>
> On the nightly page, we should include the UUIDs from Adium.app.dSYM. One of
> these will be in every Adium crash log, in the Binary Images section; this
> is the quickest way we can look up the correct dSYM-bundle archive for a
> given nightly. You can use mdls to obtain these UUIDs:

Honestly, this sounds like a ridiculous amount of work to look up a
crashlog. Even if we roll this into a webpage that does all the work
of lookup for us (this is a _requirement_ to adopt, I'd say: there's
no way I'm going to do this manually), it's still going to require
substantially more effort to look up a crashlog than currently just
looking at it. It's ~2mb savings on our binary size, which is
significant, but not enough for me to jump at the extra amount of
effort to see a crashlog.

Every single ticket with a crashlog, every single user on IRC with a
crashlog, is going to require converting the log from stripped to
symboled. This is a major change to dealing with crashes.

*Especially* with the nightly, with it changing every day…I think I
would never look up a crashlog's true values. We field way too many
crashlogs with Adium to have a several-step process to use them.

I'm not sure I see the benefit; I'd rather not decrease memory usage
than making supporting crashes extra work for myself. Convenience >
users' happiness.

Zac