People have asked my why Tinderbox does not build with client.mk, when we recommend that they build that way. Switching Tinderbox over to client.mk would give people more confidence in it. It would also improve build-moz-smoke.pl by separating the configuration from the script logic. As it is right now, it is difficult to update the script because of the conflicts with the configuration data. To make the switch, you will need to create a mozilla/.mozconfig. Alternatively, you can name the script anything you want and then set $MOZCONFIG to be the fullpath and name of the script. To set up a script, either look at the Build Configurator, http://cvs-mirror.mozilla.org/webtools/build/config.cgi or ask me to set up the options you want.

i've asked briano about this; he's going to keep his ports pages building manually, but i'm going to convert the linux tinderbox to use client.mk.

I think we should use client.mk and NOT --with-nspr everywhere. If there are problems with either of these, then we should fix those. Doing a manual hack on a per-platform basis seems contrary to the universal portability doctrine. But, if the excuse is inertia (it already is working) then I understand.

So, today we have a case where this would have helped. There is an error in configure. However, the tinderbox script goes on to the rest of the build anyway. This makes it more difficult to track down the problem.

how would that change if we were using client.mk instead of configuring manually?

client.mk will kill the whole build if configure died (Of course, you could also add this feature to your current tinderbox script). This way you see the error on Tinderbox immediately rather than waiting for a build to finish (one that will probably fail anyway).

just setting a target milestone to get it off the radar of bugs with no milestones

this is done for the main tinderbox pages

I can vouch for that. Marking verified.